{
  "id": 148561,
  "title": "External Data Thread",
  "url": "/competitions/global-wheat-detection/discussion/148561",
  "author_name": "Maggie",
  "post_date": "2020-05-04T19:55:12.546000",
  "votes": 5,
  "comment_count": 104,
  "views": 0,
  "content": "<p>Per the competition rules, post links to your external data sources here before the deadline specified. Once it has been posted, you do not need to post it again.</p>",
  "messages": [
    {
      "id": 947309,
      "postDate": "2020-07-27T07:20:19.460Z",
      "content": "<p>some nice and clear pictures we took in our hometown Wuhan and in another beautiful city Paris, feel free to use them.\n<a href=\"https://www.kaggle.com/meliao/no-wheat-image\">https://www.kaggle.com/meliao/no-wheat-image</a></p>",
      "rawMarkdown": "some nice and clear pictures we took in our hometown Wuhan and in another beautiful city Paris, feel free to use them.\nhttps://www.kaggle.com/meliao/no-wheat-image",
      "votes": 6
    },
    {
      "id": 859526,
      "postDate": "2020-05-24T14:22:17.173Z",
      "content": "<p><a href=\"/maggiemd\">@maggiemd</a> I'm a little confused as to what constitutes \"external data sources\" here. I notice other Kagglers posting model weights and python packages. Do these have to be reported? (I've checked the competition rules and it wasn't clear to me) Thank you</p>",
      "rawMarkdown": "@maggiemd I'm a little confused as to what constitutes \"external data sources\" here. I notice other Kagglers posting model weights and python packages. Do these have to be reported? (I've checked the competition rules and it wasn't clear to me) Thank you",
      "votes": 6,
      "replies": [
        {
          "id": 902535,
          "postDate": "2020-06-26T07:56:24.543Z",
          "content": "<p><a href=\"/bendvd\">@bendvd</a> I noticed in another thread you said you'd be providing a clear update in a few days. So in this side I think it might be useful to get some clarity too please:</p>\n\n<ul>\n<li><p>Does data have to be shared if we created it ourselves? (by whatever method whether automatic or totally manual)</p></li>\n<li><p>Is python code that's not sitting directly in our kernel, but rather as a \"data source\", considered an external data source that has to be shared? That questions considered in the following cases: \n(a) we wrote it ourselves from scratch\n(b) it's already totally public (for me, I think it doesn't make sense to have to share this - what if I have a really novel approach using public code, and in the act of sharing it I give away my edge)\n(c) it's already totally public, we forked it and modified it.\nAnd in any of the cases above, if the rules insist that we do share it here, then what if we literally copy paste it all in our kernel rather than have it as a data source. Does the rule still apply?</p></li>\n<li><p>Further to the above, considering the case in which the rules insist that we do share. What if I take a shared dataset like the efficientdet repo for instance, and copy it into a private data for myself. Am I breaking the rules? What if I tweak a few lines of code in it and keep it private?</p></li>\n<li><p>Are weights considered datasets? Do we have to share them here? What if I have a novel approach, and in the act of sharing the weights I give it away to everyone. (Like let's pretend cyclegan were useful in this competition but only I knew it - by sharing pretrained weights I'd tip everyone off to my master plan).</p></li>\n<li><p>It wasn't clear to me what the sharing deadline was. Maybe I missed it, could you please let me know.</p></li>\n</ul>\n\n<p>I know that's a lot, but really if we don't have clear answers to it I think it's a big blocker.</p>",
          "rawMarkdown": "@bendvd I noticed in another thread you said you'd be providing a clear update in a few days. So in this side I think it might be useful to get some clarity too please:\n\n- Does data have to be shared if we created it ourselves? (by whatever method whether automatic or totally manual)\n\n- Is python code that's not sitting directly in our kernel, but rather as a \"data source\", considered an external data source that has to be shared? That questions considered in the following cases: \n    (a) we wrote it ourselves from scratch\n    (b) it's already totally public (for me, I think it doesn't make sense to have to share this - what if I have a really novel approach using public code, and in the act of sharing it I give away my edge)\n    (c) it's already totally public, we forked it and modified it.\n  And in any of the cases above, if the rules insist that we do share it here, then what if we literally copy paste it all in our kernel rather than have it as a data source. Does the rule still apply?\n\n- Further to the above, considering the case in which the rules insist that we do share. What if I take a shared dataset like the efficientdet repo for instance, and copy it into a private data for myself. Am I breaking the rules? What if I tweak a few lines of code in it and keep it private?\n\n- Are weights considered datasets? Do we have to share them here? What if I have a novel approach, and in the act of sharing the weights I give it away to everyone. (Like let's pretend cyclegan were useful in this competition but only I knew it - by sharing pretrained weights I'd tip everyone off to my master plan).\n\n- It wasn't clear to me what the sharing deadline was. Maybe I missed it, could you please let me know.\n\nI know that's a lot, but really if we don't have clear answers to it I think it's a big blocker.",
          "votes": 1
        },
        {
          "id": 903106,
          "postDate": "2020-06-26T15:18:22.460Z",
          "content": "<p>Thanks for asking these, they are great, important questions that I wish were more clearly laid out in an FAQ. The only one I know the answer to: the deadline for sharing external datasets is the entry deadline on the timeline.</p>",
          "rawMarkdown": "Thanks for asking these, they are great, important questions that I wish were more clearly laid out in an FAQ. The only one I know the answer to: the deadline for sharing external datasets is the entry deadline on the timeline.",
          "votes": 1
        },
        {
          "id": 911332,
          "postDate": "2020-07-01T17:00:20.480Z",
          "content": "<p>Thanks for your questions, and thanks to <a href=\"/matthewmasters\">@matthewmasters</a> for pointing me here. Sorry for the delay in answering. The answers, as they apply to this competition, are as follows:</p>\n\n<ol>\n<li>If you’re making changes to the training set - for instance, updating annotations - no, you don’t need to share that. If it’s an entirely new dataset, though - yes, you do need to share it.</li>\n<li>No. Utility scripts are part of your code and don’t need to be shared. You do need to share any external data they use.</li>\n<li>No. As long as the external dataset is publicly declared, you can put it in a private dataset and use it in your code.</li>\n<li>If you created the pretrained weights then no, you don’t need to share them. They’re your own output. If they’re pretrained weights you downloaded from somewhere, then they’re an external dataset and need to be declared.</li>\n<li>The sharing deadline is 1 week before competition close.</li>\n</ol>",
          "rawMarkdown": "Thanks for your questions, and thanks to @matthewmasters for pointing me here. Sorry for the delay in answering. The answers, as they apply to this competition, are as follows:\n\n1. If you’re making changes to the training set - for instance, updating annotations - no, you don’t need to share that. If it’s an entirely new dataset, though - yes, you do need to share it.\n2. No. Utility scripts are part of your code and don’t need to be shared. You do need to share any external data they use.\n3. No. As long as the external dataset is publicly declared, you can put it in a private dataset and use it in your code.\n4. If you created the pretrained weights then no, you don’t need to share them. They’re your own output. If they’re pretrained weights you downloaded from somewhere, then they’re an external dataset and need to be declared.\n5. The sharing deadline is 1 week before competition close.",
          "votes": 3
        },
        {
          "id": 911360,
          "postDate": "2020-07-01T17:14:33.567Z",
          "content": "<p><a href=\"/philculliton\">@philculliton</a> Thank you for taking the time to answer these! It's appreciated :)</p>",
          "rawMarkdown": "@philculliton Thank you for taking the time to answer these! It's appreciated :)",
          "votes": 2
        },
        {
          "id": 911513,
          "postDate": "2020-07-01T19:38:47.060Z",
          "content": "<p><a href=\"/philculliton\">@philculliton</a> awesome thank you for the response</p>",
          "rawMarkdown": "@philculliton awesome thank you for the response",
          "votes": 2
        },
        {
          "id": 930458,
          "postDate": "2020-07-15T13:29:23.697Z",
          "content": "<blockquote>\n  <p>If you created the pretrained weights then no, you don’t need to share them.</p>\n</blockquote>\n\n<p><a href=\"/philculliton\">@philculliton</a>, so the notebook only needs to contain inference and I can train my network on my own machine and just upload network weights?</p>",
          "rawMarkdown": "&gt; If you created the pretrained weights then no, you don’t need to share them.\n\n@philculliton, so the notebook only needs to contain inference and I can train my network on my own machine and just upload network weights?"
        }
      ]
    },
    {
      "id": 833427,
      "postDate": "2020-05-04T19:55:12.547Z",
      "content": "<p>Per the competition rules, post links to your external data sources here before the deadline specified. Once it has been posted, you do not need to post it again.</p>",
      "rawMarkdown": "Per the competition rules, post links to your external data sources here before the deadline specified. Once it has been posted, you do not need to post it again.",
      "votes": 5
    },
    {
      "id": 836376,
      "postDate": "2020-05-06T23:50:41.733Z",
      "content": "<p>I'm looking to use an external dataset to expand the number of negative samples. The best thing I've found so far is\n<a href=\"https://www.kaggle.com/ravirajsinh45/crop-and-weed-detection-data-with-bounding-boxes#agri_0_1017.txt\">https://www.kaggle.com/ravirajsinh45/crop-and-weed-detection-data-with-bounding-boxes#agri_0_1017.txt</a>\nthough I'm not too confident this will actually be of any use. If anyone else has datasets in mind for this task, I'd appreciate hearing about it.</p>",
      "rawMarkdown": "I'm looking to use an external dataset to expand the number of negative samples. The best thing I've found so far is\nhttps://www.kaggle.com/ravirajsinh45/crop-and-weed-detection-data-with-bounding-boxes#agri_0_1017.txt\nthough I'm not too confident this will actually be of any use. If anyone else has datasets in mind for this task, I'd appreciate hearing about it.",
      "votes": 3
    },
    {
      "id": 955649,
      "postDate": "2020-08-02T19:21:38.147Z",
      "content": "<p><a href=\"/maggiemd\">@maggiemd</a> Are optimizers and schedulers part of the External Data Thread and need to be reported? Have not seen that kind of post before in any other competition, like for AdamW,Radam,schedulers,dataloaders that are not built-in, so it's good to know, even if it's to late this time to post/report.</p>",
      "rawMarkdown": "@maggiemd Are optimizers and schedulers part of the External Data Thread and need to be reported? Have not seen that kind of post before in any other competition, like for AdamW,Radam,schedulers,dataloaders that are not built-in, so it's good to know, even if it's to late this time to post/report.",
      "votes": 1,
      "replies": [
        {
          "id": 956269,
          "postDate": "2020-08-03T11:32:41.407Z",
          "content": "<p><a href=\"/kirderf\">@kirderf</a> I have the same doubt. Did you get your answer? If not it will be a good option to create a new discussion topic as I do not see any response here in the past few days neither from Kaggle team or the organisers. </p>",
          "rawMarkdown": "@kirderf I have the same doubt. Did you get your answer? If not it will be a good option to create a new discussion topic as I do not see any response here in the past few days neither from Kaggle team or the organisers. "
        }
      ]
    },
    {
      "id": 951191,
      "postDate": "2020-07-30T01:23:24.213Z",
      "content": "<p>Hello,If i use the dataset someone public in Kaggle,should I post it?</p>",
      "rawMarkdown": "Hello,If i use the dataset someone public in Kaggle,should I post it?",
      "votes": 1,
      "replies": [
        {
          "id": 954934,
          "postDate": "2020-08-02T07:54:42.270Z",
          "content": "<p>I have the same question. If it needs to be posted then it's already late as the external data thread has already been closed. \n<a href=\"/maggiemd\">@maggiemd</a> <a href=\"/bendvd\">@bendvd</a> \nEdit: Tagging Maggie and David for better reach.</p>",
          "rawMarkdown": "I have the same question. If it needs to be posted then it's already late as the external data thread has already been closed. \n@maggiemd @bendvd \nEdit: Tagging Maggie and David for better reach."
        }
      ]
    },
    {
      "id": 950446,
      "postDate": "2020-07-29T12:00:06.850Z",
      "content": "<p>I may be using these</p>\n\n<p><a href=\"https://ai.googleblog.com/2020/05/open-sourcing-bit-exploring-large-scale.html\">https://ai.googleblog.com/2020/05/open-sourcing-bit-exploring-large-scale.html</a>\nModels from PyTorch hub</p>",
      "rawMarkdown": "I may be using these\n\nhttps://ai.googleblog.com/2020/05/open-sourcing-bit-exploring-large-scale.html\nModels from PyTorch hub",
      "votes": 1,
      "replies": [
        {
          "id": 950999,
          "postDate": "2020-07-29T19:34:50.450Z",
          "content": "<p>The external data thread closed yesterday by the way. So if no one else posted the \"BiT\" pretrained weights, I don't think you can use them</p>",
          "rawMarkdown": "The external data thread closed yesterday by the way. So if no one else posted the \"BiT\" pretrained weights, I don't think you can use them",
          "votes": 1
        },
        {
          "id": 951378,
          "postDate": "2020-07-30T05:30:30.137Z",
          "content": "<p>Oh OK.No problem I do not think I will have time to use it.</p>",
          "rawMarkdown": "Oh OK.No problem I do not think I will have time to use it."
        }
      ]
    },
    {
      "id": 950265,
      "postDate": "2020-07-29T09:58:41.213Z",
      "content": "<p>Hello,can I process the train.csv and use this instead of the original one in the notebook I submitted?Thanks</p>",
      "rawMarkdown": "Hello,can I process the train.csv and use this instead of the original one in the notebook I submitted?Thanks",
      "votes": 1,
      "replies": [
        {
          "id": 950994,
          "postDate": "2020-07-29T19:30:21.940Z",
          "content": "<p>Yes</p>",
          "rawMarkdown": "Yes"
        },
        {
          "id": 951185,
          "postDate": "2020-07-30T01:18:23.713Z",
          "content": "<p>Thanks,and YOLOv5 can not use.Is it right?</p>",
          "rawMarkdown": "Thanks,and YOLOv5 can not use.Is it right?"
        },
        {
          "id": 951313,
          "postDate": "2020-07-30T04:01:18.193Z",
          "content": "<p>Correct. Yolov5 is liscneced GPL3 and therefore is not allowed in this competition.</p>",
          "rawMarkdown": "Correct. Yolov5 is liscneced GPL3 and therefore is not allowed in this competition.",
          "votes": 1
        }
      ]
    },
    {
      "id": 948237,
      "postDate": "2020-07-27T18:51:58.777Z",
      "content": "<p><a href=\"https://zenodo.org/record/1471626#.W9H_XFUzZjU\">Wheat Ear Dataset </a></p>",
      "rawMarkdown": "[Wheat Ear Dataset ](https://zenodo.org/record/1471626#.W9H_XFUzZjU)",
      "votes": 1,
      "replies": [
        {
          "id": 948282,
          "postDate": "2020-07-27T19:24:43.840Z",
          "content": "<p>Did this help?</p>",
          "rawMarkdown": "Did this help?"
        },
        {
          "id": 951019,
          "postDate": "2020-07-29T20:06:15.400Z",
          "content": "<p>These images seem to be original full-size images of the source \"Arvalis_1\" in the training dataset. </p>",
          "rawMarkdown": "These images seem to be original full-size images of the source \"Arvalis_1\" in the training dataset. ",
          "votes": 3
        },
        {
          "id": 951021,
          "postDate": "2020-07-29T20:08:22.913Z",
          "content": "<p>Oh interesting, I was looking at them and was surprised how close they were to the images in the training set. That would make sense :)</p>",
          "rawMarkdown": "Oh interesting, I was looking at them and was surprised how close they were to the images in the training set. That would make sense :)"
        },
        {
          "id": 954627,
          "postDate": "2020-08-01T23:30:43.277Z",
          "content": "<p>can we use this dataset in this competion?</p>",
          "rawMarkdown": "can we use this dataset in this competion?",
          "votes": 1
        },
        {
          "id": 954632,
          "postDate": "2020-08-01T23:50:07.567Z",
          "content": "<p><a href=\"/nvnnghia\">@nvnnghia</a> I would think so. Why not?</p>",
          "rawMarkdown": "@nvnnghia I would think so. Why not?",
          "votes": 1
        },
        {
          "id": 954635,
          "postDate": "2020-08-02T00:09:34.020Z",
          "content": "<p><a href=\"/matthewmasters\">@matthewmasters</a> Thanks for your answer, I am worry about license. </p>",
          "rawMarkdown": "@matthewmasters Thanks for your answer, I am worry about license. ",
          "votes": 1
        },
        {
          "id": 954708,
          "postDate": "2020-08-02T02:37:36.090Z",
          "content": "<p>Oh, I did not notice the Creative Commons licence. I'm honestly not sure</p>",
          "rawMarkdown": "Oh, I did not notice the Creative Commons licence. I'm honestly not sure",
          "votes": 1
        },
        {
          "id": 954715,
          "postDate": "2020-08-02T02:49:03.073Z",
          "content": "<p>Thanks. </p>",
          "rawMarkdown": "Thanks. "
        }
      ]
    },
    {
      "id": 872875,
      "postDate": "2020-06-03T15:18:09.057Z",
      "content": "<p>The pre-trained models from GluonCV:</p>\n\n<ul>\n<li><a href=\"https://gluon-cv.mxnet.io/model_zoo/detection.html\">https://gluon-cv.mxnet.io/model_zoo/detection.html</a></li>\n</ul>\n\n<p>Weighted-Boxes-Fusion:</p>\n\n<ul>\n<li><a href=\"https://github.com/ZFTurbo/Weighted-Boxes-Fusion\">https://github.com/ZFTurbo/Weighted-Boxes-Fusion</a></li>\n</ul>",
      "rawMarkdown": "The pre-trained models from GluonCV:\n\n* [https://gluon-cv.mxnet.io/model_zoo/detection.html](https://gluon-cv.mxnet.io/model_zoo/detection.html)\n\nWeighted-Boxes-Fusion:\n\n* [https://github.com/ZFTurbo/Weighted-Boxes-Fusion](https://github.com/ZFTurbo/Weighted-Boxes-Fusion)",
      "votes": 1
    },
    {
      "id": 949765,
      "postDate": "2020-07-28T22:36:37.143Z",
      "content": "<p><a href=\"https://github.com/open-mmlab/mmdetection\">https://github.com/open-mmlab/mmdetection</a>\n<a href=\"https://github.com/VDIGPKU/CBNet\">https://github.com/VDIGPKU/CBNet</a>\n<a href=\"https://github.com/JialeCao001/D2Det\">https://github.com/JialeCao001/D2Det</a>\n<a href=\"https://github.com/shinya7y/UniverseNet\">https://github.com/shinya7y/UniverseNet</a>\n<a href=\"https://github.com/thangvubk/Cascade-RPN\">https://github.com/thangvubk/Cascade-RPN</a>\n<a href=\"https://github.com/WXinlong/SOLO\">https://github.com/WXinlong/SOLO</a>\n<a href=\"https://github.com/Sense-X/TSD\">https://github.com/Sense-X/TSD</a>\n<a href=\"https://github.com/rwightman/efficientdet-pytorch\">https://github.com/rwightman/efficientdet-pytorch</a>\n<a href=\"https://github.com/rwightman/pytorch-image-models\">https://github.com/rwightman/pytorch-image-models</a>\n<a href=\"https://github.com/zylo117/Yet-Another-EfficientDet-Pytorch\">https://github.com/zylo117/Yet-Another-EfficientDet-Pytorch</a>\n<a href=\"https://github.com/joe-siyuan-qiao/DetectoRS\">https://github.com/joe-siyuan-qiao/DetectoRS</a>\n<a href=\"https://github.com/facebookresearch/detectron2\">https://github.com/facebookresearch/detectron2</a>\n<a href=\"https://github.com/TuSimple/simpledet\">https://github.com/TuSimple/simpledet</a>\n<a href=\"https://github.com/naoto0804/pytorch-AdaIN\">https://github.com/naoto0804/pytorch-AdaIN</a>\n<a href=\"https://github.com/eriklindernoren/PyTorch-GAN\">https://github.com/eriklindernoren/PyTorch-GAN</a>\n<a href=\"https://github.com/NVlabs/stylegan2\">https://github.com/NVlabs/stylegan2</a>\n<a href=\"https://github.com/clovaai/stargan-v2\">https://github.com/clovaai/stargan-v2</a>\n<a href=\"https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix\">https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix</a></p>",
      "rawMarkdown": "\nhttps://github.com/open-mmlab/mmdetection\nhttps://github.com/VDIGPKU/CBNet\nhttps://github.com/JialeCao001/D2Det\nhttps://github.com/shinya7y/UniverseNet\nhttps://github.com/thangvubk/Cascade-RPN\nhttps://github.com/WXinlong/SOLO\nhttps://github.com/Sense-X/TSD\nhttps://github.com/rwightman/efficientdet-pytorch\nhttps://github.com/rwightman/pytorch-image-models\nhttps://github.com/zylo117/Yet-Another-EfficientDet-Pytorch\nhttps://github.com/joe-siyuan-qiao/DetectoRS\nhttps://github.com/facebookresearch/detectron2\nhttps://github.com/TuSimple/simpledet\nhttps://github.com/naoto0804/pytorch-AdaIN\nhttps://github.com/eriklindernoren/PyTorch-GAN\nhttps://github.com/NVlabs/stylegan2\nhttps://github.com/clovaai/stargan-v2\nhttps://github.com/junyanz/pytorch-CycleGAN-and-pix2pix"
    },
    {
      "id": 853968,
      "postDate": "2020-05-19T16:34:48.983Z",
      "content": "<p><a href=\"https://github.com/rwightman/efficientdet-pytorch\">https://github.com/rwightman/efficientdet-pytorch</a>\n<a href=\"https://github.com/rwightman/pytorch-image-models\">https://github.com/rwightman/pytorch-image-models</a>\n<a href=\"https://www.kaggle.com/greatgamedota/timm-package\">https://www.kaggle.com/greatgamedota/timm-package</a>\n<a href=\"https://www.kaggle.com/kaushal2896/pycocotools\">https://www.kaggle.com/kaushal2896/pycocotools</a>\n<a href=\"https://www.kaggle.com/shonenkov/omegaconf\">https://www.kaggle.com/shonenkov/omegaconf</a>\n<a href=\"https://www.kaggle.com/mathurinache/efficientdet\">https://www.kaggle.com/mathurinache/efficientdet</a></p>",
      "rawMarkdown": "https://github.com/rwightman/efficientdet-pytorch\nhttps://github.com/rwightman/pytorch-image-models\nhttps://www.kaggle.com/greatgamedota/timm-package\nhttps://www.kaggle.com/kaushal2896/pycocotools\nhttps://www.kaggle.com/shonenkov/omegaconf\nhttps://www.kaggle.com/mathurinache/efficientdet",
      "votes": -1,
      "replies": [
        {
          "id": 858896,
          "postDate": "2020-05-23T23:13:44.213Z",
          "content": "<p>Please post the original external links, not just your own/or other's kernel repackaging. \n<a href=\"https://github.com/rwightman/efficientdet-pytorch\">https://github.com/rwightman/efficientdet-pytorch</a>\n<a href=\"https://github.com/rwightman/pytorch-image-models\">https://github.com/rwightman/pytorch-image-models</a></p>\n\n<p>As the author of those packages I contribute with a permissive license so anyone can use them in applications like this. But as the author I would also like to have other users come back to the original source -- so that they have access to the latest versions, can provide feedback, bug reports, contribute via PR, etc. For a similar reason, a link to the original source in your kernels would be much appreciated. It is in fact required by Apache license terms to include some form of attribution for a redistribution of the code in documentation or file. Thanks!</p>",
          "rawMarkdown": "Please post the original external links, not just your own/or other's kernel repackaging. \nhttps://github.com/rwightman/efficientdet-pytorch\nhttps://github.com/rwightman/pytorch-image-models\n\nAs the author of those packages I contribute with a permissive license so anyone can use them in applications like this. But as the author I would also like to have other users come back to the original source -- so that they have access to the latest versions, can provide feedback, bug reports, contribute via PR, etc. For a similar reason, a link to the original source in your kernels would be much appreciated. It is in fact required by Apache license terms to include some form of attribution for a redistribution of the code in documentation or file. Thanks!",
          "votes": 26
        },
        {
          "id": 862679,
          "postDate": "2020-05-26T18:28:06.803Z",
          "content": "<p><a href=\"/rwightman\">@rwightman</a> Sure! I have added links on your repositories. Thank you for your work! </p>",
          "rawMarkdown": "@rwightman Sure! I have added links on your repositories. Thank you for your work! ",
          "votes": 2
        },
        {
          "id": 862805,
          "postDate": "2020-05-26T20:52:59.453Z",
          "content": "<p><a href=\"/shonenkov\">@shonenkov</a> thanks, glad to see it's working well.. I've recently forwarded your train kernel link to a few who've asked about training on alternate datasets with different classes and augmentation pipelines as it's a concise example. I'll link it in the README if that sounds good with you?</p>",
          "rawMarkdown": "@shonenkov thanks, glad to see it's working well.. I've recently forwarded your train kernel link to a few who've asked about training on alternate datasets with different classes and augmentation pipelines as it's a concise example. I'll link it in the README if that sounds good with you?",
          "votes": 2
        },
        {
          "id": 862826,
          "postDate": "2020-05-26T21:21:43.307Z",
          "content": "<p><a href=\"/rwightman\">@rwightman</a> It is really good idea! I think Global Wheat Detection is good example for training with custom datasets using your pytorch effdet implementations. It can be very useful for other users. Thank you a lot!</p>",
          "rawMarkdown": "@rwightman It is really good idea! I think Global Wheat Detection is good example for training with custom datasets using your pytorch effdet implementations. It can be very useful for other users. Thank you a lot!"
        }
      ]
    },
    {
      "id": 923137,
      "postDate": "2020-07-10T14:56:38.453Z",
      "content": "<p>Hi I am new, This comment is Non-Sense. Just to one up my tier.</p>",
      "rawMarkdown": "Hi I am new, This comment is Non-Sense. Just to one up my tier.\n",
      "votes": -13
    },
    {
      "id": 958128,
      "postDate": "2020-08-04T19:42:09.433Z",
      "content": "<p><strong>We Are using these sources:</strong></p>\n\n<p><a href=\"https://github.com/rwightman/efficientdet-pytorch\">https://github.com/rwightman/efficientdet-pytorch</a></p>\n\n<p><a href=\"https://www.kaggle.com/shonenkov/timm-efficientdet-pytorch\">https://www.kaggle.com/shonenkov/timm-efficientdet-pytorch</a></p>\n\n<p><a href=\"https://www.kaggle.com/shonenkov/weightedboxesfusion\">https://www.kaggle.com/shonenkov/weightedboxesfusion</a></p>\n\n<p><a href=\"https://www.kaggle.com/hawkey/yolov5train\">https://www.kaggle.com/hawkey/yolov5train</a> \nOriginal Source:  <a href=\"https://github.com/ultralytics/yolov5\">https://github.com/ultralytics/yolov5</a></p>\n\n<p><strong>Configuration Files</strong>\n<a href=\"https://www.kaggle.com/orkatz2/configyolo5\">https://www.kaggle.com/orkatz2/configyolo5</a>\n<a href=\"https://www.kaggle.com/mihirmavalankar/omegaconf\">https://www.kaggle.com/mihirmavalankar/omegaconf</a>\n<a href=\"https://www.kaggle.com/shonenkov/omegaconf\">https://www.kaggle.com/shonenkov/omegaconf</a></p>\n\n<p><strong>Weights</strong>\n<a href=\"https://www.kaggle.com/jaychakalsiya/global-wheat-frcnn-23-06\">https://www.kaggle.com/jaychakalsiya/global-wheat-frcnn-23-06</a>\n<a href=\"https://www.kaggle.com/mihirmavalankar/detrmodels\">https://www.kaggle.com/mihirmavalankar/detrmodels</a>\n<a href=\"https://www.kaggle.com/mihirmavalankar/retinanetmodels\">https://www.kaggle.com/mihirmavalankar/retinanetmodels</a></p>\n\n<p>** Our Repo**\n<a href=\"https://github.com/jay-chakalasiya/Global-Wheat-Detction-Kaggle-2020.git\">https://github.com/jay-chakalasiya/Global-Wheat-Detction-Kaggle-2020.git</a>\nSources:\n<a href=\"https://github.com/zylo117/Yet-Another-EfficientDet-Pytorch\">https://github.com/zylo117/Yet-Another-EfficientDet-Pytorch</a>\n<a href=\"https://github.com/facebookresearch/detr\">https://github.com/facebookresearch/detr</a>\n<a href=\"https://github.com/yhenon/pytorch-retinanet\">https://github.com/yhenon/pytorch-retinanet</a>\n<a href=\"https://github.com/ultralytics/yolov5\">https://github.com/ultralytics/yolov5</a>\n<a href=\"https://github.com/ZFTurbo/Weighted-Boxes-Fusion\">https://github.com/ZFTurbo/Weighted-Boxes-Fusion</a></p>\n\n<p><em>Please comment if I'm forgetting to mention any of the original Git references,\nThank You</em></p>",
      "rawMarkdown": "**We Are using these sources:**\n\nhttps://github.com/rwightman/efficientdet-pytorch\n\nhttps://www.kaggle.com/shonenkov/timm-efficientdet-pytorch\n\nhttps://www.kaggle.com/shonenkov/weightedboxesfusion\n\nhttps://www.kaggle.com/hawkey/yolov5train \nOriginal Source:  https://github.com/ultralytics/yolov5\n\n**Configuration Files**\nhttps://www.kaggle.com/orkatz2/configyolo5\nhttps://www.kaggle.com/mihirmavalankar/omegaconf\nhttps://www.kaggle.com/shonenkov/omegaconf\n\n**Weights**\nhttps://www.kaggle.com/jaychakalsiya/global-wheat-frcnn-23-06\nhttps://www.kaggle.com/mihirmavalankar/detrmodels\nhttps://www.kaggle.com/mihirmavalankar/retinanetmodels\n\n** Our Repo**\nhttps://github.com/jay-chakalasiya/Global-Wheat-Detction-Kaggle-2020.git\nSources:\nhttps://github.com/zylo117/Yet-Another-EfficientDet-Pytorch\nhttps://github.com/facebookresearch/detr\nhttps://github.com/yhenon/pytorch-retinanet\nhttps://github.com/ultralytics/yolov5\nhttps://github.com/ZFTurbo/Weighted-Boxes-Fusion\n\n\n*Please comment if I'm forgetting to mention any of the original Git references,\nThank You*\n",
      "votes": -1,
      "replies": [
        {
          "id": 958133,
          "postDate": "2020-08-04T19:46:39.340Z",
          "content": "<p>Just so you know, reporting external data is due before the entry deadline, not the final deadline. Luckily it doesn't look like you're using anything that hasn't been posted here already :)</p>",
          "rawMarkdown": "Just so you know, reporting external data is due before the entry deadline, not the final deadline. Luckily it doesn't look like you're using anything that hasn't been posted here already :)",
          "votes": 2
        }
      ]
    },
    {
      "id": 958904,
      "postDate": "2020-08-05T07:57:06.390Z",
      "content": "<p><a href=\"https://www.kaggle.com/shonenkov/inference-efficientdet\">https://www.kaggle.com/shonenkov/inference-efficientdet</a></p>\n\n<p><a href=\"https://www.kaggle.com/linyangsdu/detectors-wheat-inference-public\">https://www.kaggle.com/linyangsdu/detectors-wheat-inference-public</a></p>",
      "rawMarkdown": "https://www.kaggle.com/shonenkov/inference-efficientdet\n\nhttps://www.kaggle.com/linyangsdu/detectors-wheat-inference-public"
    },
    {
      "id": 950107,
      "postDate": "2020-07-29T07:27:08.223Z",
      "content": "<p>I am using/ experimenting with these pytorch pretrained weights as backbone, \n<code>\nmodel_urls = {\n    'fasterrcnn_resnet50_fpn_coco': 'https://download.pytorch.org/models/fasterrcnn_resnet50_fpn_coco-258fb6c6.pth',\n     'resnet18': 'https://download.pytorch.org/models/resnet18-5c106cde.pth',\n    'resnet34': 'https://download.pytorch.org/models/resnet34-333f7ec4.pth',\n    'resnet50': 'https://download.pytorch.org/models/resnet50-19c8e357.pth',\n    'resnet101': 'https://download.pytorch.org/models/resnet101-5d3b4d8f.pth',\n    'resnet152': 'https://download.pytorch.org/models/resnet152-b121ed2d.pth',\n    'resnext50_32x4d': 'https://download.pytorch.org/models/resnext50_32x4d-7cdf4587.pth',\n    'resnext101_32x8d': 'https://download.pytorch.org/models/resnext101_32x8d-8ba56ff5.pth',\n    'wide_resnet50_2': 'https://download.pytorch.org/models/wide_resnet50_2-95faca4d.pth',\n    'wide_resnet101_2': 'https://download.pytorch.org/models/wide_resnet101_2-32ee1156.pth',\n}\n</code></p>",
      "rawMarkdown": "I am using/ experimenting with these pytorch pretrained weights as backbone, \n```\nmodel_urls = {\n    'fasterrcnn_resnet50_fpn_coco': 'https://download.pytorch.org/models/fasterrcnn_resnet50_fpn_coco-258fb6c6.pth',\n     'resnet18': 'https://download.pytorch.org/models/resnet18-5c106cde.pth',\n    'resnet34': 'https://download.pytorch.org/models/resnet34-333f7ec4.pth',\n    'resnet50': 'https://download.pytorch.org/models/resnet50-19c8e357.pth',\n    'resnet101': 'https://download.pytorch.org/models/resnet101-5d3b4d8f.pth',\n    'resnet152': 'https://download.pytorch.org/models/resnet152-b121ed2d.pth',\n    'resnext50_32x4d': 'https://download.pytorch.org/models/resnext50_32x4d-7cdf4587.pth',\n    'resnext101_32x8d': 'https://download.pytorch.org/models/resnext101_32x8d-8ba56ff5.pth',\n    'wide_resnet50_2': 'https://download.pytorch.org/models/wide_resnet50_2-95faca4d.pth',\n    'wide_resnet101_2': 'https://download.pytorch.org/models/wide_resnet101_2-32ee1156.pth',\n}\n```"
    },
    {
      "id": 949711,
      "postDate": "2020-07-28T20:41:39.450Z",
      "content": "<p><a href=\"https://www.kaggle.com/angqx95/effdet-pytorch\">https://www.kaggle.com/angqx95/effdet-pytorch</a>\n<a href=\"https://www.kaggle.com/angqx95/efficientdet-model\">https://www.kaggle.com/angqx95/efficientdet-model</a>\n<a href=\"https://www.kaggle.com/shonenkov/omegaconf\">https://www.kaggle.com/shonenkov/omegaconf</a>\n<a href=\"https://www.kaggle.com/kaushal2896/pycocotools\">https://www.kaggle.com/kaushal2896/pycocotools</a>\n<a href=\"https://www.kaggle.com/angqx95/torch-img-model\">https://www.kaggle.com/angqx95/torch-img-model</a>\n<a href=\"https://www.kaggle.com/shonenkov/weightedboxesfusion\">https://www.kaggle.com/shonenkov/weightedboxesfusion</a></p>",
      "rawMarkdown": "https://www.kaggle.com/angqx95/effdet-pytorch\nhttps://www.kaggle.com/angqx95/efficientdet-model\nhttps://www.kaggle.com/shonenkov/omegaconf\nhttps://www.kaggle.com/kaushal2896/pycocotools\nhttps://www.kaggle.com/angqx95/torch-img-model\nhttps://www.kaggle.com/shonenkov/weightedboxesfusion"
    },
    {
      "id": 949351,
      "postDate": "2020-07-28T15:24:04.703Z",
      "content": "<p>Tried some of these below, but no results so far.</p>\n\n<p><strong>Few keras/tensorflow references for object detection :</strong>\n<a href=\"https://github.com/google/automl/tree/master/efficientdet\">https://github.com/google/automl/tree/master/efficientdet</a>\n<a href=\"https://github.com/fizyr/keras-retinanet\">https://github.com/fizyr/keras-retinanet</a> and <a href=\"https://github.com/fizyr/tf-retinanet\">https://github.com/fizyr/tf-retinanet</a>\n<a href=\"https://github.com/xuannianz/EfficientDet\">https://github.com/xuannianz/EfficientDet</a>\n<a href=\"https://github.com/calmisential/EfficientDet_TensorFlow2\">https://github.com/calmisential/EfficientDet_TensorFlow2</a></p>\n\n<p>*<em>Pre trained Model files : *</em>\n<a href=\"https://github.com/tensorflow/\">https://github.com/tensorflow/</a>\n<a href=\"https://github.com/qubvel/efficientnet\">https://github.com/qubvel/efficientnet</a>\n<a href=\"https://github.com/Callidior/keras-applications/releases\">https://github.com/Callidior/keras-applications/releases</a></p>",
      "rawMarkdown": "Tried some of these below, but no results so far.\n\n**Few keras/tensorflow references for object detection :**\nhttps://github.com/google/automl/tree/master/efficientdet\nhttps://github.com/fizyr/keras-retinanet and https://github.com/fizyr/tf-retinanet\nhttps://github.com/xuannianz/EfficientDet\nhttps://github.com/calmisential/EfficientDet_TensorFlow2\n\n**Pre trained Model files : **\nhttps://github.com/tensorflow/\nhttps://github.com/qubvel/efficientnet\nhttps://github.com/Callidior/keras-applications/releases\n\n"
    },
    {
      "id": 948451,
      "postDate": "2020-07-28T00:51:00.690Z",
      "content": "<p>Thanks <a href=\"/whurobin\">@whurobin</a> \n<a href=\"https://github.com/wuxinwang1997/wheatdetection\">https://github.com/wuxinwang1997/wheatdetection</a>\n<a href=\"https://www.kaggle.com/whurobin/wbf-over-tta-single-model-fasterrcnn-resnest\">https://www.kaggle.com/whurobin/wbf-over-tta-single-model-fasterrcnn-resnest</a></p>",
      "rawMarkdown": "Thanks @whurobin \nhttps://github.com/wuxinwang1997/wheatdetection\nhttps://www.kaggle.com/whurobin/wbf-over-tta-single-model-fasterrcnn-resnest"
    },
    {
      "id": 947040,
      "postDate": "2020-07-27T03:22:14.920Z",
      "content": "<p>keras-centernet\n<a href=\"https://github.com/see--/keras-centernet\">https://github.com/see--/keras-centernet</a></p>",
      "rawMarkdown": "keras-centernet\n[https://github.com/see--/keras-centernet](https://github.com/see--/keras-centernet)"
    },
    {
      "id": 946410,
      "postDate": "2020-07-26T15:01:16.553Z",
      "content": "<p>Spike wheat dataset <a href=\"https://www.kaggle.com/voglinio/spike-wheat-dataset\">https://www.kaggle.com/voglinio/spike-wheat-dataset</a></p>",
      "rawMarkdown": "Spike wheat dataset https://www.kaggle.com/voglinio/spike-wheat-dataset",
      "replies": [
        {
          "id": 946709,
          "postDate": "2020-07-26T19:02:07.107Z",
          "content": "<p>I'm curious if you saw any improvement with it? I noticed the labels are pretty noisy and the photos are taken from a side view instead of top-down</p>",
          "rawMarkdown": "I'm curious if you saw any improvement with it? I noticed the labels are pretty noisy and the photos are taken from a side view instead of top-down"
        },
        {
          "id": 947828,
          "postDate": "2020-07-27T14:00:21.690Z",
          "content": "<p>Can we use this dataset? no license is similar MIT license?(I don't know license in details.)\n<a href=\"https://www.kaggle.com/c/global-wheat-detection/discussion/164346#917316\">https://www.kaggle.com/c/global-wheat-detection/discussion/164346#917316</a></p>",
          "rawMarkdown": "Can we use this dataset? no license is similar MIT license?(I don't know license in details.)\nhttps://www.kaggle.com/c/global-wheat-detection/discussion/164346#917316"
        },
        {
          "id": 947989,
          "postDate": "2020-07-27T15:38:36.313Z",
          "content": "<p>Not that much of improvement.. and since it is a little fuzzy whether we could use it I didn't push it that much. </p>",
          "rawMarkdown": "Not that much of improvement.. and since it is a little fuzzy whether we could use it I didn't push it that much. "
        }
      ]
    },
    {
      "id": 946200,
      "postDate": "2020-07-26T12:34:21.223Z",
      "content": "<p>PP-YOLO with PaddleDetection\n<a href=\"https://github.com/PaddlePaddle/PaddleDetection\">https://github.com/PaddlePaddle/PaddleDetection</a>\n<a href=\"https://arxiv.org/abs/2007.12099\">https://arxiv.org/abs/2007.12099</a></p>",
      "rawMarkdown": "PP-YOLO with PaddleDetection\nhttps://github.com/PaddlePaddle/PaddleDetection\nhttps://arxiv.org/abs/2007.12099"
    },
    {
      "id": 943338,
      "postDate": "2020-07-24T09:54:38.833Z",
      "content": "<p>External data:\n<a href=\"https://www.kaggle.com/ravirajsinh45/crop-and-weed-detection-data-with-bounding-boxes\">https://www.kaggle.com/ravirajsinh45/crop-and-weed-detection-data-with-bounding-boxes</a>?</p>",
      "rawMarkdown": "External data:\nhttps://www.kaggle.com/ravirajsinh45/crop-and-weed-detection-data-with-bounding-boxes?"
    },
    {
      "id": 941434,
      "postDate": "2020-07-23T08:07:40.673Z",
      "content": "<p>torchvision pretrained models: <a href=\"https://pytorch.org/docs/stable/torchvision/models.html\">https://pytorch.org/docs/stable/torchvision/models.html</a>\nCOCO dataset: <a href=\"https://cocodataset.org/\">https://cocodataset.org/</a></p>",
      "rawMarkdown": "torchvision pretrained models: https://pytorch.org/docs/stable/torchvision/models.html\nCOCO dataset: https://cocodataset.org/"
    },
    {
      "id": 937434,
      "postDate": "2020-07-21T02:49:04.020Z",
      "content": "<p><a href=\"https://github.com/ZFTurbo/Weighted-Boxes-Fusion/tree/v1.0.5\" target=\"_blank\">https://github.com/ZFTurbo/Weighted-Boxes-Fusion/tree/v1.0.5</a><br>\n<a href=\"https://github.com/NVIDIA/apex\" target=\"_blank\">https://github.com/NVIDIA/apex</a><br>\n<a href=\"https://github.com/ultralytics/yolov5\" target=\"_blank\">https://github.com/ultralytics/yolov5</a></p>",
      "rawMarkdown": "https://github.com/ZFTurbo/Weighted-Boxes-Fusion/tree/v1.0.5\nhttps://github.com/NVIDIA/apex\nhttps://github.com/ultralytics/yolov5\n"
    },
    {
      "id": 937195,
      "postDate": "2020-07-20T19:45:23.940Z",
      "content": "<p><a href=\"https://github.com/rwightman/efficientdet-pytorch\">https://github.com/rwightman/efficientdet-pytorch</a>\n<a href=\"https://github.com/rwightman/pytorch-image-models\">https://github.com/rwightman/pytorch-image-models</a>\n<a href=\"https://github.com/ZFTurbo/Weighted-Boxes-Fusion\">https://github.com/ZFTurbo/Weighted-Boxes-Fusion</a>\n<a href=\"https://www.kaggle.com/greatgamedota/timm-package\">https://www.kaggle.com/greatgamedota/timm-package</a>\n<a href=\"https://www.kaggle.com/kaushal2896/pycocotools\">https://www.kaggle.com/kaushal2896/pycocotools</a>\n<a href=\"https://www.kaggle.com/shonenkov/omegaconf\">https://www.kaggle.com/shonenkov/omegaconf</a>\n<a href=\"https://www.kaggle.com/shonenkov/weightedboxesfusion\">https://www.kaggle.com/shonenkov/weightedboxesfusion</a>\n<a href=\"https://www.kaggle.com/shonenkov/timm-efficientdet-pytorch\">https://www.kaggle.com/shonenkov/timm-efficientdet-pytorch</a>\n<a href=\"https://www.kaggle.com/mathurinache/efficientdet\">https://www.kaggle.com/mathurinache/efficientdet</a></p>",
      "rawMarkdown": "https://github.com/rwightman/efficientdet-pytorch\nhttps://github.com/rwightman/pytorch-image-models\nhttps://github.com/ZFTurbo/Weighted-Boxes-Fusion\nhttps://www.kaggle.com/greatgamedota/timm-package\nhttps://www.kaggle.com/kaushal2896/pycocotools\nhttps://www.kaggle.com/shonenkov/omegaconf\nhttps://www.kaggle.com/shonenkov/weightedboxesfusion\nhttps://www.kaggle.com/shonenkov/timm-efficientdet-pytorch\nhttps://www.kaggle.com/mathurinache/efficientdet"
    },
    {
      "id": 936666,
      "postDate": "2020-07-20T12:30:20.423Z",
      "content": "<p><a href=\"https://github.com/NVIDIA/apex\" target=\"_blank\">https://github.com/NVIDIA/apex</a><br>\n<a href=\"https://www.kaggle.com/ii5m0k3ii/nvidia-apex\" target=\"_blank\">https://www.kaggle.com/ii5m0k3ii/nvidia-apex</a></p>",
      "rawMarkdown": "https://github.com/NVIDIA/apex\nhttps://www.kaggle.com/ii5m0k3ii/nvidia-apex"
    },
    {
      "id": 936134,
      "postDate": "2020-07-20T02:36:34.987Z",
      "content": "<p><a href=\"https://github.com/rwightman/efficientdet-pytorch\">https://github.com/rwightman/efficientdet-pytorch</a>\n<a href=\"https://github.com/rwightman/pytorch-image-models\">https://github.com/rwightman/pytorch-image-models</a>\n<a href=\"https://www.kaggle.com/greatgamedota/timm-package\">https://www.kaggle.com/greatgamedota/timm-package</a>\n<a href=\"https://www.kaggle.com/kaushal2896/pycocotools\">https://www.kaggle.com/kaushal2896/pycocotools</a>\n<a href=\"https://www.kaggle.com/shonenkov/omegaconf\">https://www.kaggle.com/shonenkov/omegaconf</a>\n<a href=\"https://www.kaggle.com/mathurinache/efficientdet\">https://www.kaggle.com/mathurinache/efficientdet</a></p>",
      "rawMarkdown": "https://github.com/rwightman/efficientdet-pytorch\nhttps://github.com/rwightman/pytorch-image-models\nhttps://www.kaggle.com/greatgamedota/timm-package\nhttps://www.kaggle.com/kaushal2896/pycocotools\nhttps://www.kaggle.com/shonenkov/omegaconf\nhttps://www.kaggle.com/mathurinache/efficientdet"
    },
    {
      "id": 935701,
      "postDate": "2020-07-19T15:15:12.213Z",
      "content": "<p>Spike Dataset:<br>\n<a href=\"https://sourceforge.net/projects/spike-dataset/\" target=\"_blank\">https://sourceforge.net/projects/spike-dataset/</a></p>\n<p>YOLOv4:</p>\n<ul>\n<li><a href=\"https://github.com/AlexeyAB/darknet\" target=\"_blank\">https://github.com/AlexeyAB/darknet</a></li>\n</ul>\n<p>EfficientDet:</p>\n<ul>\n<li><a href=\"https://github.com/rwightman/efficientdet-pytorch\" target=\"_blank\">https://github.com/rwightman/efficientdet-pytorch</a></li>\n<li><a href=\"https://github.com/rwightman/pytorch-image-models\" target=\"_blank\">https://github.com/rwightman/pytorch-image-models</a></li>\n<li><a href=\"https://github.com/google/automl/tree/master/efficientdet\" target=\"_blank\">https://github.com/google/automl/tree/master/efficientdet</a></li>\n<li><a href=\"https://github.com/zylo117/Yet-Another-EfficientDet-Pytorch\" target=\"_blank\">https://github.com/zylo117/Yet-Another-EfficientDet-Pytorch</a></li>\n</ul>\n<p>RetinaNet:</p>\n<ul>\n<li><a href=\"https://github.com/yhenon/pytorch-retinanet\" target=\"_blank\">https://github.com/yhenon/pytorch-retinanet</a></li>\n</ul>\n<p>CenterNet:</p>\n<ul>\n<li><a href=\"https://github.com/xingyizhou/CenterNet\" target=\"_blank\">https://github.com/xingyizhou/CenterNet</a></li>\n</ul>\n<p>ResNeSt:</p>\n<ul>\n<li><a href=\"https://github.com/zhanghang1989/detectron2-ResNeSt\" target=\"_blank\">https://github.com/zhanghang1989/detectron2-ResNeSt</a></li>\n</ul>",
      "rawMarkdown": "Spike Dataset:\nhttps://sourceforge.net/projects/spike-dataset/\n\nYOLOv4:\n* https://github.com/AlexeyAB/darknet\n\nEfficientDet:\n* https://github.com/rwightman/efficientdet-pytorch\n* https://github.com/rwightman/pytorch-image-models\n* https://github.com/google/automl/tree/master/efficientdet\n* https://github.com/zylo117/Yet-Another-EfficientDet-Pytorch\n\nRetinaNet:\n* https://github.com/yhenon/pytorch-retinanet\n\nCenterNet:\n* https://github.com/xingyizhou/CenterNet\n\nResNeSt:\n* https://github.com/zhanghang1989/detectron2-ResNeSt"
    },
    {
      "id": 931291,
      "postDate": "2020-07-16T05:29:21.170Z",
      "content": "<p>Can we use youtube videos to weakly supervise our models?\ne.g. facebook uses Instagram photos to get better performance.</p>\n\n<p>Maybe this isn't allowed because of the very strict license issues but seems like a interesting approach (for CVPRR).</p>\n\n<p>some wheat videos\n<a href=\"https://www.youtube.com/watch?v=YTix0yymVWM\">Agriculture Expert Guide to Increase Wheat Production. Wheat production in Pakistan</a></p>",
      "rawMarkdown": "Can we use youtube videos to weakly supervise our models?\ne.g. facebook uses Instagram photos to get better performance.\n\nMaybe this isn't allowed because of the very strict license issues but seems like a interesting approach (for CVPRR).\n\nsome wheat videos\n[Agriculture Expert Guide to Increase Wheat Production. Wheat production in Pakistan](https://www.youtube.com/watch?v=YTix0yymVWM)\n\n",
      "replies": [
        {
          "id": 931303,
          "postDate": "2020-07-16T05:36:28.593Z",
          "content": "<p>Not sure what you mean exactly. The video you shared (and many on the same channel) does not contain any video of wheat itself, just people discussing wheat agriculture.</p>\n\n<p>Personally, I would refrain from using any youtube content for the competition. Even if it is licensed properly there have been big issues in the past with this (<a href=\"https://www.kaggle.com/c/deepfake-detection-challenge/discussion/157983\">source</a>).</p>\n\n<p>But if you are able to come up with something for CVPRR, I think it's a great idea!</p>",
          "rawMarkdown": "Not sure what you mean exactly. The video you shared (and many on the same channel) does not contain any video of wheat itself, just people discussing wheat agriculture.\n\nPersonally, I would refrain from using any youtube content for the competition. Even if it is licensed properly there have been big issues in the past with this ([source](https://www.kaggle.com/c/deepfake-detection-challenge/discussion/157983)).\n\nBut if you are able to come up with something for CVPRR, I think it's a great idea!"
        }
      ]
    },
    {
      "id": 921474,
      "postDate": "2020-07-09T10:20:02.417Z",
      "content": "<p>I know that yolov5 has issues.\nAnd I'm not sure how to share it, but I leave a post in this thread.</p>\n\n<p><a href=\"https://github.com/ultralytics/yolov5\">https://github.com/ultralytics/yolov5</a>\n<a href=\"https://www.kaggle.com/shonenkov/weightedboxesfusion\">https://www.kaggle.com/shonenkov/weightedboxesfusion</a>\n<a href=\"https://github.com/rwightman/efficientdet-pytorch\">https://github.com/rwightman/efficientdet-pytorch</a>\n<a href=\"https://github.com/rwightman/pytorch-image-models\">https://github.com/rwightman/pytorch-image-models</a>\n<a href=\"https://github.com/zylo117/Yet-Another-EfficientDet-Pytorch\">https://github.com/zylo117/Yet-Another-EfficientDet-Pytorch</a>\n<a href=\"https://github.com/open-mmlab/mmdetection\">https://github.com/open-mmlab/mmdetection</a>\n<a href=\"https://github.com/facebookresearch/detr/\">https://github.com/facebookresearch/detr/</a></p>",
      "rawMarkdown": "I know that yolov5 has issues.\nAnd I'm not sure how to share it, but I leave a post in this thread.\n\n[https://github.com/ultralytics/yolov5](https://github.com/ultralytics/yolov5)\n[https://www.kaggle.com/shonenkov/weightedboxesfusion](https://www.kaggle.com/shonenkov/weightedboxesfusion)\n[https://github.com/rwightman/efficientdet-pytorch](https://github.com/rwightman/efficientdet-pytorch)\n[https://github.com/rwightman/pytorch-image-models](https://github.com/rwightman/pytorch-image-models)\n[https://github.com/zylo117/Yet-Another-EfficientDet-Pytorch](https://github.com/zylo117/Yet-Another-EfficientDet-Pytorch)\n[https://github.com/open-mmlab/mmdetection](https://github.com/open-mmlab/mmdetection)\n[https://github.com/facebookresearch/detr/](https://github.com/facebookresearch/detr/)"
    },
    {
      "id": 918083,
      "postDate": "2020-07-07T01:07:42.593Z",
      "content": "<p>\"Once it has been posted, you do not need to post it again.\" Does that mean if someone posted YOLOv5 everyone doesn't have to post it? I'm asking because I'm seeing people posting the same things over and over. I just want to be clear if we have to post our own external data or ?? </p>",
      "rawMarkdown": "\"Once it has been posted, you do not need to post it again.\" Does that mean if someone posted YOLOv5 everyone doesn't have to post it? I'm asking because I'm seeing people posting the same things over and over. I just want to be clear if we have to post our own external data or ?? ",
      "replies": [
        {
          "id": 918106,
          "postDate": "2020-07-07T01:55:01.180Z",
          "content": "<p>That's the idea: \"*<em>Y</em>*ou *<em>O</em>*nly *<em>P</em>*ost *<em>O</em>*nce\" in this discussion thread.</p>\n\n<p>Others need to just skim through this before posting any <em>new</em> external data.</p>",
          "rawMarkdown": "That's the idea: \"**Y**ou **O**nly **P**ost **O**nce\" in this discussion thread.\n\nOthers need to just skim through this before posting any *new* external data.",
          "votes": 2
        }
      ]
    },
    {
      "id": 916803,
      "postDate": "2020-07-06T03:46:55.407Z",
      "content": "<p>Pretrained models using 400K synthetic data: \n<a href=\"https://csiro-robotics.github.io/UPGen-Webpage/\">https://csiro-robotics.github.io/UPGen-Webpage/</a></p>",
      "rawMarkdown": "Pretrained models using 400K synthetic data: \nhttps://csiro-robotics.github.io/UPGen-Webpage/\n\n",
      "replies": [
        {
          "id": 916852,
          "postDate": "2020-07-06T04:46:01Z",
          "content": "<p>I'm not sure that's allowed since it's released under a non-commercial license.</p>",
          "rawMarkdown": "I'm not sure that's allowed since it's released under a non-commercial license."
        }
      ]
    },
    {
      "id": 916274,
      "postDate": "2020-07-05T14:21:59.270Z",
      "content": "<p>What about Synthetic data? Is that considered as external data?</p>",
      "rawMarkdown": "What about Synthetic data? Is that considered as external data?",
      "replies": [
        {
          "id": 917540,
          "postDate": "2020-07-06T15:45:47.073Z",
          "content": "<p>Can you clarify? Are you pulling this from an external source? </p>",
          "rawMarkdown": "Can you clarify? Are you pulling this from an external source? "
        },
        {
          "id": 917850,
          "postDate": "2020-07-06T19:10:44.170Z",
          "content": "<p>Nope, using GANS to produce synthetic data. It's trained using competition data.</p>",
          "rawMarkdown": "Nope, using GANS to produce synthetic data. It's trained using competition data."
        },
        {
          "id": 919750,
          "postDate": "2020-07-08T04:57:58.713Z",
          "content": "<p>lol</p>",
          "rawMarkdown": "lol",
          "votes": -2
        },
        {
          "id": 919794,
          "postDate": "2020-07-08T05:36:05.633Z",
          "content": "<p>The requirement is for external data, not data that you created yourself. </p>",
          "rawMarkdown": "The requirement is for external data, not data that you created yourself. ",
          "votes": 1
        },
        {
          "id": 920584,
          "postDate": "2020-07-08T17:35:14.077Z",
          "content": "<p>Thank you for clarifying. New here so had this doubt.  </p>",
          "rawMarkdown": "Thank you for clarifying. New here so had this doubt.  ",
          "votes": 1
        },
        {
          "id": 927017,
          "postDate": "2020-07-13T06:41:43.603Z",
          "content": "<p><a href=\"/maggiemd\">@maggiemd</a> Just to clarify .. let's say I make some wheat pictures my self and annotate them. Would that be considered 'external data' that needs to be shared? GAN generated images are OK ... so I'm guessing that would be OK too, but just want to be 100% certain. </p>",
          "rawMarkdown": "@maggiemd Just to clarify .. let's say I make some wheat pictures my self and annotate them. Would that be considered 'external data' that needs to be shared? GAN generated images are OK ... so I'm guessing that would be OK too, but just want to be 100% certain. "
        },
        {
          "id": 927075,
          "postDate": "2020-07-13T07:24:42.460Z",
          "content": "<p>What do you mean by \"make some wheat pictures my self\"? Like, you're actually taking these images yourself, or you are finding them online somewhere?</p>",
          "rawMarkdown": "What do you mean by \"make some wheat pictures my self\"? Like, you're actually taking these images yourself, or you are finding them online somewhere?"
        },
        {
          "id": 929309,
          "postDate": "2020-07-14T15:45:41.537Z",
          "content": "<p>Making my self. Hypothetically, I'm literally taking a camera, taking some photos in a wheat field and adding a few bounding boxes. </p>",
          "rawMarkdown": "Making my self. Hypothetically, I'm literally taking a camera, taking some photos in a wheat field and adding a few bounding boxes. "
        }
      ]
    },
    {
      "id": 915096,
      "postDate": "2020-07-04T13:19:47.593Z",
      "content": "<p>model -\n<a href=\"https://github.com/ultralytics/yolov5\">https://github.com/ultralytics/yolov5</a>\nWBM-\n<a href=\"https://www.kaggle.com/shonenkov/weightedboxesfusion\">https://www.kaggle.com/shonenkov/weightedboxesfusion</a></p>",
      "rawMarkdown": "model -\nhttps://github.com/ultralytics/yolov5\nWBM-\nhttps://www.kaggle.com/shonenkov/weightedboxesfusion"
    },
    {
      "id": 914818,
      "postDate": "2020-07-04T09:00:24.647Z",
      "content": "<p>model -\n<a href=\"https://github.com/ultralytics/yolov5\">https://github.com/ultralytics/yolov5</a>\nWBM-\n<a href=\"https://www.kaggle.com/shonenkov/weightedboxesfusion\">https://www.kaggle.com/shonenkov/weightedboxesfusion</a></p>",
      "rawMarkdown": "model -\nhttps://github.com/ultralytics/yolov5\nWBM-\nhttps://www.kaggle.com/shonenkov/weightedboxesfusion"
    },
    {
      "id": 909935,
      "postDate": "2020-07-01T00:08:41.930Z",
      "content": "<p>RESNEXT WSL\n<a href=\"https://pytorch.org/hub/facebookresearch_WSL-Images_resnext/\">https://pytorch.org/hub/facebookresearch_WSL-Images_resnext/</a></p>",
      "rawMarkdown": "RESNEXT WSL\nhttps://pytorch.org/hub/facebookresearch_WSL-Images_resnext/"
    },
    {
      "id": 901619,
      "postDate": "2020-06-25T15:51:47.467Z",
      "content": "<p><a href=\"https://www.kaggle.com/imazekishota/yolov5-output-conf\">https://www.kaggle.com/imazekishota/yolov5-output-conf</a></p>",
      "rawMarkdown": "https://www.kaggle.com/imazekishota/yolov5-output-conf"
    },
    {
      "id": 899121,
      "postDate": "2020-06-24T02:12:21.780Z",
      "content": "<p><a href=\"https://github.com/tensorflow/models\">https://github.com/tensorflow/models</a></p>",
      "rawMarkdown": "https://github.com/tensorflow/models"
    },
    {
      "id": 899068,
      "postDate": "2020-06-24T00:20:42.397Z",
      "content": "<p><a href=\"https://www.kaggle.com/orkatz2/yolov5-train\">https://www.kaggle.com/orkatz2/yolov5-train</a></p>",
      "rawMarkdown": "https://www.kaggle.com/orkatz2/yolov5-train"
    },
    {
      "id": 899062,
      "postDate": "2020-06-24T00:09:16.540Z",
      "content": "<p>Weighted boxes fusion:\n<a href=\"https://www.kaggle.com/shonenkov/weightedboxesfusion\">https://www.kaggle.com/shonenkov/weightedboxesfusion</a></p>",
      "rawMarkdown": "Weighted boxes fusion:\nhttps://www.kaggle.com/shonenkov/weightedboxesfusion"
    },
    {
      "id": 899054,
      "postDate": "2020-06-23T23:55:47.407Z",
      "content": "<p>YOLOv5:\n<a href=\"https://github.com/ultralytics/yolov5\">https://github.com/ultralytics/yolov5</a></p>",
      "rawMarkdown": "YOLOv5:\nhttps://github.com/ultralytics/yolov5"
    },
    {
      "id": 881871,
      "postDate": "2020-06-11T12:30:54.147Z",
      "content": "<p><a href=\"/maggiemd\">@maggiemd</a> Quick question about external data - I see that we are not allowed to use custom packages, but that we are allowed to use external data. Does external data include open-source .py class files? </p>\n\n<p>If we can upload .py files, at what point does it start to be considered a 'custom package'; as long as it's not compiled it's fine?</p>",
      "rawMarkdown": "@maggiemd Quick question about external data - I see that we are not allowed to use custom packages, but that we are allowed to use external data. Does external data include open-source .py class files? \n\nIf we can upload .py files, at what point does it start to be considered a 'custom package'; as long as it's not compiled it's fine?",
      "replies": [
        {
          "id": 882073,
          "postDate": "2020-06-11T15:08:36.283Z",
          "content": "<p><a href=\"/tim1515\">@tim1515</a> could I ask where it says that we can't use custom packages please?</p>",
          "rawMarkdown": "@tim1515 could I ask where it says that we can't use custom packages please?"
        },
        {
          "id": 882123,
          "postDate": "2020-06-11T15:45:45.563Z",
          "content": "<p>On the <a href=\"https://www.kaggle.com/c/global-wheat-detection/overview/code-requirements\">Code Requirements</a> page it states \"No custom packages enabled in kernels\"</p>",
          "rawMarkdown": "On the [Code Requirements](https://www.kaggle.com/c/global-wheat-detection/overview/code-requirements) page it states \"No custom packages enabled in kernels\""
        },
        {
          "id": 882190,
          "postDate": "2020-06-11T16:30:28.843Z",
          "content": "<p><a href=\"/matthewmasters\">@matthewmasters</a> thanks!\nWeird... so with this EfficientDet implementation we can't use the repo as a dataset? If not, then I'll just end up copy pasting the whole thing in the kernel. Or am I not understanding something.</p>",
          "rawMarkdown": "@matthewmasters thanks!\nWeird... so with this EfficientDet implementation we can't use the repo as a dataset? If not, then I'll just end up copy pasting the whole thing in the kernel. Or am I not understanding something."
        },
        {
          "id": 882215,
          "postDate": "2020-06-11T16:45:00.350Z",
          "content": "<p>I believe custom packages are allowed in the form of datasets, but not through pip or anything like that because there’s no internet access, though I’m not 100% on these rules. It would be nice if Kaggle could give us some insight there but they don’t seem to monitor these threads much. </p>\n\n<p>I might go back and look at the ARC competition as the same rule was in place there and I remember some discussion about it</p>",
          "rawMarkdown": "I believe custom packages are allowed in the form of datasets, but not through pip or anything like that because there’s no internet access, though I’m not 100% on these rules. It would be nice if Kaggle could give us some insight there but they don’t seem to monitor these threads much. \n\nI might go back and look at the ARC competition as the same rule was in place there and I remember some discussion about it"
        },
        {
          "id": 882275,
          "postDate": "2020-06-11T17:16:34.113Z",
          "content": "<p>I searched around there and still couldn't find a direct answer. One Kaggle employee said that they removed that from the Code Requirements because it is redundant with the No internet requirement.</p>\n\n<p>I am fairly certain using datasets for external code is allowed.</p>",
          "rawMarkdown": "I searched around there and still couldn't find a direct answer. One Kaggle employee said that they removed that from the Code Requirements because it is redundant with the No internet requirement.\n\nI am fairly certain using datasets for external code is allowed.",
          "votes": 2
        },
        {
          "id": 882315,
          "postDate": "2020-06-11T18:00:45.730Z",
          "content": "<p>Thanks for looking at it. I suppose if I wind up with a good LB position I'll do a proper check</p>",
          "rawMarkdown": "Thanks for looking at it. I suppose if I wind up with a good LB position I'll do a proper check",
          "votes": 2
        },
        {
          "id": 882505,
          "postDate": "2020-06-11T21:29:03.703Z",
          "content": "<p>I'll probably do the same thing then, just uploading to the datasets - that saves having a crazy long notebook. Thanks for the feedback all!</p>",
          "rawMarkdown": "I'll probably do the same thing then, just uploading to the datasets - that saves having a crazy long notebook. Thanks for the feedback all!"
        },
        {
          "id": 888650,
          "postDate": "2020-06-16T13:36:45.733Z",
          "content": "<p>If the external code is openly available, not a problem (i.e to organize your notebook). It would be a problem if it is private code that can't be share. The winning solutions are supposed to be open-source.</p>",
          "rawMarkdown": "If the external code is openly available, not a problem (i.e to organize your notebook). It would be a problem if it is private code that can't be share. The winning solutions are supposed to be open-source.",
          "votes": 2
        },
        {
          "id": 899327,
          "postDate": "2020-06-24T07:00:25.563Z",
          "content": "<p><a href=\"/bendvd\">@bendvd</a> what if it is private code that we are willing to share? So if I want to take some code and put it in a py file but it belongs to my private solution, but then once the competition is over I open access to it.</p>",
          "rawMarkdown": "@bendvd what if it is private code that we are willing to share? So if I want to take some code and put it in a py file but it belongs to my private solution, but then once the competition is over I open access to it.",
          "votes": 2
        }
      ]
    },
    {
      "id": 863159,
      "postDate": "2020-05-27T05:37:19.287Z",
      "content": "<p><a href=\"https://plantimages.nottingham.ac.uk/\">https://plantimages.nottingham.ac.uk/</a>\n<a href=\"https://pytorch.org/docs/stable/torchvision/models.html\">https://pytorch.org/docs/stable/torchvision/models.html</a></p>",
      "rawMarkdown": "https://plantimages.nottingham.ac.uk/\nhttps://pytorch.org/docs/stable/torchvision/models.html",
      "replies": [
        {
          "id": 863222,
          "postDate": "2020-05-27T06:51:30.907Z",
          "content": "<p>Do you actually think that dataset will be of use? All the images are taken in an unnatural setting (single plants with black backdrop) and are taken from a side view instead of the nadir view that the images in this dataset use. I just think it's too far from the target distribution to be useful in anyway.</p>",
          "rawMarkdown": "Do you actually think that dataset will be of use? All the images are taken in an unnatural setting (single plants with black backdrop) and are taken from a side view instead of the nadir view that the images in this dataset use. I just think it's too far from the target distribution to be useful in anyway."
        },
        {
          "id": 864294,
          "postDate": "2020-05-28T00:01:16.810Z",
          "content": "<p>Actually haven't tried it yet..\nJust bringing it up so that I don't forget lol</p>",
          "rawMarkdown": "Actually haven't tried it yet..\nJust bringing it up so that I don't forget lol"
        },
        {
          "id": 879104,
          "postDate": "2020-06-09T07:36:49.307Z",
          "content": "<p>:)</p>",
          "rawMarkdown": ":)",
          "votes": -1
        }
      ]
    },
    {
      "id": 859493,
      "postDate": "2020-05-24T13:37:08.300Z",
      "content": "<p><a href=\"https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix\">https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix</a></p>",
      "rawMarkdown": "https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix\n"
    },
    {
      "id": 851855,
      "postDate": "2020-05-18T01:05:33.437Z",
      "content": "<ol>\n<li><a href=\"https://github.com/zylo117/Yet-Another-EfficientDet-Pytorch\">https://github.com/zylo117/Yet-Another-EfficientDet-Pytorch</a></li>\n<li><a href=\"https://github.com/open-mmlab/mmdetection\">https://github.com/open-mmlab/mmdetection</a></li>\n</ol>",
      "rawMarkdown": "1. https://github.com/zylo117/Yet-Another-EfficientDet-Pytorch\n2. https://github.com/open-mmlab/mmdetection"
    },
    {
      "id": 850092,
      "postDate": "2020-05-16T10:36:06.560Z",
      "content": "<p>Out of curiosity, what if I use pictures I made myself ? Should they be uploaded somewhere ?</p>",
      "rawMarkdown": "Out of curiosity, what if I use pictures I made myself ? Should they be uploaded somewhere ?",
      "replies": [
        {
          "id": 919792,
          "postDate": "2020-07-08T05:35:36.877Z",
          "content": "<p>The requirement is for external data, not data that you created yourself. </p>",
          "rawMarkdown": "The requirement is for external data, not data that you created yourself. "
        },
        {
          "id": 942745,
          "postDate": "2020-07-24T01:23:46.313Z",
          "content": "<p>😶 In the past, if a competitor annotated the data themselves, they've always had to share within the 7-day period. For example, if this were a medical imaging classification competition and a radiologist manually created bounding box or segmentation labels on top of the competition training data and got the leg up due to her own SME, that's always been interpreted as external data here on Kaggle. I imagine that if a farmer went and took 20000 images of different wheat stalks, that too would be considered external data right?</p>\n\n<p>On the other hand, if a programmer created a morphological algorithm that (in the first case) took care of the segmentation task in a programmatic manner, I'd understand that to fall into the \"part of their kernel's code\" category. If possible, please confirm and thank you for clarification.</p>",
          "rawMarkdown": "😶 In the past, if a competitor annotated the data themselves, they've always had to share within the 7-day period. For example, if this were a medical imaging classification competition and a radiologist manually created bounding box or segmentation labels on top of the competition training data and got the leg up due to her own SME, that's always been interpreted as external data here on Kaggle. I imagine that if a farmer went and took 20000 images of different wheat stalks, that too would be considered external data right?\n\nOn the other hand, if a programmer created a morphological algorithm that (in the first case) took care of the segmentation task in a programmatic manner, I'd understand that to fall into the \"part of their kernel's code\" category. If possible, please confirm and thank you for clarification.",
          "votes": -2
        }
      ]
    },
    {
      "id": 834373,
      "postDate": "2020-05-05T14:12:20.493Z",
      "content": "<p>Mask-RCNN (matterport) source code. \n- <a href=\"https://www.kaggle.com/ipythonx/maskrcnn-keras-source-code\">https://www.kaggle.com/ipythonx/maskrcnn-keras-source-code</a></p>\n\n<p>Backbone weights:\n- <a href=\"https://www.kaggle.com/ipythonx/keras-pretrained-imagenet-weights\">https://www.kaggle.com/ipythonx/keras-pretrained-imagenet-weights</a>\n- <a href=\"https://www.kaggle.com/ipythonx/efficientnet-keras-noisystudent-weights-b0b7\">https://www.kaggle.com/ipythonx/efficientnet-keras-noisystudent-weights-b0b7</a>\n- <a href=\"https://www.kaggle.com/ipythonx/cocowg\">https://www.kaggle.com/ipythonx/cocowg</a></p>",
      "rawMarkdown": "Mask-RCNN (matterport) source code. \n- https://www.kaggle.com/ipythonx/maskrcnn-keras-source-code\n\nBackbone weights:\n- https://www.kaggle.com/ipythonx/keras-pretrained-imagenet-weights\n- https://www.kaggle.com/ipythonx/efficientnet-keras-noisystudent-weights-b0b7\n- https://www.kaggle.com/ipythonx/cocowg",
      "replies": [
        {
          "id": 845596,
          "postDate": "2020-05-13T10:02:47.830Z",
          "content": "<p>Are weights considered external data that we have to report? cc <a href=\"/maggiemd\">@maggiemd</a> </p>",
          "rawMarkdown": "Are weights considered external data that we have to report? cc @maggiemd "
        },
        {
          "id": 853692,
          "postDate": "2020-05-19T12:06:49.457Z",
          "content": "<p>I suppose if you use ImageNet or other open-source pre-trained weight than not if you use data that is not public and of wheat heads, I assume the data should suffice.</p>",
          "rawMarkdown": "I suppose if you use ImageNet or other open-source pre-trained weight than not if you use data that is not public and of wheat heads, I assume the data should suffice."
        }
      ]
    },
    {
      "id": 927147,
      "postDate": "2020-07-13T08:21:48.627Z",
      "rawMarkdown": "",
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 947309,
      "author_name": "Alex_Wang",
      "author_url": "",
      "post_date": "2020-07-27T07:20:19.460000",
      "content": "<p>some nice and clear pictures we took in our hometown Wuhan and in another beautiful city Paris, feel free to use them.\n<a href=\"https://www.kaggle.com/meliao/no-wheat-image\">https://www.kaggle.com/meliao/no-wheat-image</a></p>",
      "votes": 6,
      "replies": []
    },
    {
      "id": 859526,
      "author_name": "Alexander Soare",
      "author_url": "",
      "post_date": "2020-05-24T14:22:17.173000",
      "content": "<p><a href=\"/maggiemd\">@maggiemd</a> I'm a little confused as to what constitutes \"external data sources\" here. I notice other Kagglers posting model weights and python packages. Do these have to be reported? (I've checked the competition rules and it wasn't clear to me) Thank you</p>",
      "votes": 6,
      "replies": [
        {
          "id": 902535,
          "author_name": "Alexander Soare",
          "author_url": "",
          "post_date": "2020-06-26T07:56:24.543000",
          "content": "<p><a href=\"/bendvd\">@bendvd</a> I noticed in another thread you said you'd be providing a clear update in a few days. So in this side I think it might be useful to get some clarity too please:</p>\n\n<ul>\n<li><p>Does data have to be shared if we created it ourselves? (by whatever method whether automatic or totally manual)</p></li>\n<li><p>Is python code that's not sitting directly in our kernel, but rather as a \"data source\", considered an external data source that has to be shared? That questions considered in the following cases: \n(a) we wrote it ourselves from scratch\n(b) it's already totally public (for me, I think it doesn't make sense to have to share this - what if I have a really novel approach using public code, and in the act of sharing it I give away my edge)\n(c) it's already totally public, we forked it and modified it.\nAnd in any of the cases above, if the rules insist that we do share it here, then what if we literally copy paste it all in our kernel rather than have it as a data source. Does the rule still apply?</p></li>\n<li><p>Further to the above, considering the case in which the rules insist that we do share. What if I take a shared dataset like the efficientdet repo for instance, and copy it into a private data for myself. Am I breaking the rules? What if I tweak a few lines of code in it and keep it private?</p></li>\n<li><p>Are weights considered datasets? Do we have to share them here? What if I have a novel approach, and in the act of sharing the weights I give it away to everyone. (Like let's pretend cyclegan were useful in this competition but only I knew it - by sharing pretrained weights I'd tip everyone off to my master plan).</p></li>\n<li><p>It wasn't clear to me what the sharing deadline was. Maybe I missed it, could you please let me know.</p></li>\n</ul>\n\n<p>I know that's a lot, but really if we don't have clear answers to it I think it's a big blocker.</p>",
          "votes": 1,
          "replies": []
        },
        {
          "id": 903106,
          "author_name": "Matt",
          "author_url": "",
          "post_date": "2020-06-26T15:18:22.460000",
          "content": "<p>Thanks for asking these, they are great, important questions that I wish were more clearly laid out in an FAQ. The only one I know the answer to: the deadline for sharing external datasets is the entry deadline on the timeline.</p>",
          "votes": 1,
          "replies": []
        },
        {
          "id": 911332,
          "author_name": "Phil Culliton",
          "author_url": "",
          "post_date": "2020-07-01T17:00:20.480000",
          "content": "<p>Thanks for your questions, and thanks to <a href=\"/matthewmasters\">@matthewmasters</a> for pointing me here. Sorry for the delay in answering. The answers, as they apply to this competition, are as follows:</p>\n\n<ol>\n<li>If you’re making changes to the training set - for instance, updating annotations - no, you don’t need to share that. If it’s an entirely new dataset, though - yes, you do need to share it.</li>\n<li>No. Utility scripts are part of your code and don’t need to be shared. You do need to share any external data they use.</li>\n<li>No. As long as the external dataset is publicly declared, you can put it in a private dataset and use it in your code.</li>\n<li>If you created the pretrained weights then no, you don’t need to share them. They’re your own output. If they’re pretrained weights you downloaded from somewhere, then they’re an external dataset and need to be declared.</li>\n<li>The sharing deadline is 1 week before competition close.</li>\n</ol>",
          "votes": 3,
          "replies": []
        },
        {
          "id": 911360,
          "author_name": "Matt",
          "author_url": "",
          "post_date": "2020-07-01T17:14:33.567000",
          "content": "<p><a href=\"/philculliton\">@philculliton</a> Thank you for taking the time to answer these! It's appreciated :)</p>",
          "votes": 2,
          "replies": []
        },
        {
          "id": 911513,
          "author_name": "Alexander Soare",
          "author_url": "",
          "post_date": "2020-07-01T19:38:47.060000",
          "content": "<p><a href=\"/philculliton\">@philculliton</a> awesome thank you for the response</p>",
          "votes": 2,
          "replies": []
        },
        {
          "id": 930458,
          "author_name": "Onix",
          "author_url": "",
          "post_date": "2020-07-15T13:29:23.697000",
          "content": "<blockquote>\n  <p>If you created the pretrained weights then no, you don’t need to share them.</p>\n</blockquote>\n\n<p><a href=\"/philculliton\">@philculliton</a>, so the notebook only needs to contain inference and I can train my network on my own machine and just upload network weights?</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 836376,
      "author_name": "Matt",
      "author_url": "",
      "post_date": "2020-05-06T23:50:41.733000",
      "content": "<p>I'm looking to use an external dataset to expand the number of negative samples. The best thing I've found so far is\n<a href=\"https://www.kaggle.com/ravirajsinh45/crop-and-weed-detection-data-with-bounding-boxes#agri_0_1017.txt\">https://www.kaggle.com/ravirajsinh45/crop-and-weed-detection-data-with-bounding-boxes#agri_0_1017.txt</a>\nthough I'm not too confident this will actually be of any use. If anyone else has datasets in mind for this task, I'd appreciate hearing about it.</p>",
      "votes": 3,
      "replies": []
    },
    {
      "id": 955649,
      "author_name": "Kirderf",
      "author_url": "",
      "post_date": "2020-08-02T19:21:38.147000",
      "content": "<p><a href=\"/maggiemd\">@maggiemd</a> Are optimizers and schedulers part of the External Data Thread and need to be reported? Have not seen that kind of post before in any other competition, like for AdamW,Radam,schedulers,dataloaders that are not built-in, so it's good to know, even if it's to late this time to post/report.</p>",
      "votes": 1,
      "replies": [
        {
          "id": 956269,
          "author_name": "bendang",
          "author_url": "",
          "post_date": "2020-08-03T11:32:41.407000",
          "content": "<p><a href=\"/kirderf\">@kirderf</a> I have the same doubt. Did you get your answer? If not it will be a good option to create a new discussion topic as I do not see any response here in the past few days neither from Kaggle team or the organisers. </p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 951191,
      "author_name": "HEEEEE",
      "author_url": "",
      "post_date": "2020-07-30T01:23:24.213000",
      "content": "<p>Hello,If i use the dataset someone public in Kaggle,should I post it?</p>",
      "votes": 1,
      "replies": [
        {
          "id": 954934,
          "author_name": "bendang",
          "author_url": "",
          "post_date": "2020-08-02T07:54:42.270000",
          "content": "<p>I have the same question. If it needs to be posted then it's already late as the external data thread has already been closed. \n<a href=\"/maggiemd\">@maggiemd</a> <a href=\"/bendvd\">@bendvd</a> \nEdit: Tagging Maggie and David for better reach.</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 950446,
      "author_name": "Vishnu Subramanian",
      "author_url": "",
      "post_date": "2020-07-29T12:00:06.850000",
      "content": "<p>I may be using these</p>\n\n<p><a href=\"https://ai.googleblog.com/2020/05/open-sourcing-bit-exploring-large-scale.html\">https://ai.googleblog.com/2020/05/open-sourcing-bit-exploring-large-scale.html</a>\nModels from PyTorch hub</p>",
      "votes": 1,
      "replies": [
        {
          "id": 950999,
          "author_name": "Matt",
          "author_url": "",
          "post_date": "2020-07-29T19:34:50.450000",
          "content": "<p>The external data thread closed yesterday by the way. So if no one else posted the \"BiT\" pretrained weights, I don't think you can use them</p>",
          "votes": 1,
          "replies": []
        },
        {
          "id": 951378,
          "author_name": "Vishnu Subramanian",
          "author_url": "",
          "post_date": "2020-07-30T05:30:30.137000",
          "content": "<p>Oh OK.No problem I do not think I will have time to use it.</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 950265,
      "author_name": "Zekun",
      "author_url": "",
      "post_date": "2020-07-29T09:58:41.213000",
      "content": "<p>Hello,can I process the train.csv and use this instead of the original one in the notebook I submitted?Thanks</p>",
      "votes": 1,
      "replies": [
        {
          "id": 950994,
          "author_name": "Matt",
          "author_url": "",
          "post_date": "2020-07-29T19:30:21.940000",
          "content": "<p>Yes</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 951185,
          "author_name": "HEEEEE",
          "author_url": "",
          "post_date": "2020-07-30T01:18:23.713000",
          "content": "<p>Thanks,and YOLOv5 can not use.Is it right?</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 951313,
          "author_name": "Stanley Zheng",
          "author_url": "",
          "post_date": "2020-07-30T04:01:18.193000",
          "content": "<p>Correct. Yolov5 is liscneced GPL3 and therefore is not allowed in this competition.</p>",
          "votes": 1,
          "replies": []
        }
      ]
    },
    {
      "id": 948237,
      "author_name": "Max",
      "author_url": "",
      "post_date": "2020-07-27T18:51:58.777000",
      "content": "<p><a href=\"https://zenodo.org/record/1471626#.W9H_XFUzZjU\">Wheat Ear Dataset </a></p>",
      "votes": 1,
      "replies": [
        {
          "id": 948282,
          "author_name": "Alexander Soare",
          "author_url": "",
          "post_date": "2020-07-27T19:24:43.840000",
          "content": "<p>Did this help?</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 951019,
          "author_name": "alexx",
          "author_url": "",
          "post_date": "2020-07-29T20:06:15.400000",
          "content": "<p>These images seem to be original full-size images of the source \"Arvalis_1\" in the training dataset. </p>",
          "votes": 3,
          "replies": []
        },
        {
          "id": 951021,
          "author_name": "Matt",
          "author_url": "",
          "post_date": "2020-07-29T20:08:22.913000",
          "content": "<p>Oh interesting, I was looking at them and was surprised how close they were to the images in the training set. That would make sense :)</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 954627,
          "author_name": "nvnn",
          "author_url": "",
          "post_date": "2020-08-01T23:30:43.277000",
          "content": "<p>can we use this dataset in this competion?</p>",
          "votes": 1,
          "replies": []
        },
        {
          "id": 954632,
          "author_name": "Matt",
          "author_url": "",
          "post_date": "2020-08-01T23:50:07.567000",
          "content": "<p><a href=\"/nvnnghia\">@nvnnghia</a> I would think so. Why not?</p>",
          "votes": 1,
          "replies": []
        },
        {
          "id": 954635,
          "author_name": "nvnn",
          "author_url": "",
          "post_date": "2020-08-02T00:09:34.020000",
          "content": "<p><a href=\"/matthewmasters\">@matthewmasters</a> Thanks for your answer, I am worry about license. </p>",
          "votes": 1,
          "replies": []
        },
        {
          "id": 954708,
          "author_name": "Matt",
          "author_url": "",
          "post_date": "2020-08-02T02:37:36.090000",
          "content": "<p>Oh, I did not notice the Creative Commons licence. I'm honestly not sure</p>",
          "votes": 1,
          "replies": []
        },
        {
          "id": 954715,
          "author_name": "nvnn",
          "author_url": "",
          "post_date": "2020-08-02T02:49:03.073000",
          "content": "<p>Thanks. </p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 872875,
      "author_name": "RangerUFO",
      "author_url": "",
      "post_date": "2020-06-03T15:18:09.057000",
      "content": "<p>The pre-trained models from GluonCV:</p>\n\n<ul>\n<li><a href=\"https://gluon-cv.mxnet.io/model_zoo/detection.html\">https://gluon-cv.mxnet.io/model_zoo/detection.html</a></li>\n</ul>\n\n<p>Weighted-Boxes-Fusion:</p>\n\n<ul>\n<li><a href=\"https://github.com/ZFTurbo/Weighted-Boxes-Fusion\">https://github.com/ZFTurbo/Weighted-Boxes-Fusion</a></li>\n</ul>",
      "votes": 1,
      "replies": []
    },
    {
      "id": 949765,
      "author_name": "Miras Amir",
      "author_url": "",
      "post_date": "2020-07-28T22:36:37.143000",
      "content": "<p><a href=\"https://github.com/open-mmlab/mmdetection\">https://github.com/open-mmlab/mmdetection</a>\n<a href=\"https://github.com/VDIGPKU/CBNet\">https://github.com/VDIGPKU/CBNet</a>\n<a href=\"https://github.com/JialeCao001/D2Det\">https://github.com/JialeCao001/D2Det</a>\n<a href=\"https://github.com/shinya7y/UniverseNet\">https://github.com/shinya7y/UniverseNet</a>\n<a href=\"https://github.com/thangvubk/Cascade-RPN\">https://github.com/thangvubk/Cascade-RPN</a>\n<a href=\"https://github.com/WXinlong/SOLO\">https://github.com/WXinlong/SOLO</a>\n<a href=\"https://github.com/Sense-X/TSD\">https://github.com/Sense-X/TSD</a>\n<a href=\"https://github.com/rwightman/efficientdet-pytorch\">https://github.com/rwightman/efficientdet-pytorch</a>\n<a href=\"https://github.com/rwightman/pytorch-image-models\">https://github.com/rwightman/pytorch-image-models</a>\n<a href=\"https://github.com/zylo117/Yet-Another-EfficientDet-Pytorch\">https://github.com/zylo117/Yet-Another-EfficientDet-Pytorch</a>\n<a href=\"https://github.com/joe-siyuan-qiao/DetectoRS\">https://github.com/joe-siyuan-qiao/DetectoRS</a>\n<a href=\"https://github.com/facebookresearch/detectron2\">https://github.com/facebookresearch/detectron2</a>\n<a href=\"https://github.com/TuSimple/simpledet\">https://github.com/TuSimple/simpledet</a>\n<a href=\"https://github.com/naoto0804/pytorch-AdaIN\">https://github.com/naoto0804/pytorch-AdaIN</a>\n<a href=\"https://github.com/eriklindernoren/PyTorch-GAN\">https://github.com/eriklindernoren/PyTorch-GAN</a>\n<a href=\"https://github.com/NVlabs/stylegan2\">https://github.com/NVlabs/stylegan2</a>\n<a href=\"https://github.com/clovaai/stargan-v2\">https://github.com/clovaai/stargan-v2</a>\n<a href=\"https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix\">https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix</a></p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 853968,
      "author_name": "Alex Shonenkov",
      "author_url": "",
      "post_date": "2020-05-19T16:34:48.983000",
      "content": "<p><a href=\"https://github.com/rwightman/efficientdet-pytorch\">https://github.com/rwightman/efficientdet-pytorch</a>\n<a href=\"https://github.com/rwightman/pytorch-image-models\">https://github.com/rwightman/pytorch-image-models</a>\n<a href=\"https://www.kaggle.com/greatgamedota/timm-package\">https://www.kaggle.com/greatgamedota/timm-package</a>\n<a href=\"https://www.kaggle.com/kaushal2896/pycocotools\">https://www.kaggle.com/kaushal2896/pycocotools</a>\n<a href=\"https://www.kaggle.com/shonenkov/omegaconf\">https://www.kaggle.com/shonenkov/omegaconf</a>\n<a href=\"https://www.kaggle.com/mathurinache/efficientdet\">https://www.kaggle.com/mathurinache/efficientdet</a></p>",
      "votes": -1,
      "replies": [
        {
          "id": 858896,
          "author_name": "RossWightman",
          "author_url": "",
          "post_date": "2020-05-23T23:13:44.213000",
          "content": "<p>Please post the original external links, not just your own/or other's kernel repackaging. \n<a href=\"https://github.com/rwightman/efficientdet-pytorch\">https://github.com/rwightman/efficientdet-pytorch</a>\n<a href=\"https://github.com/rwightman/pytorch-image-models\">https://github.com/rwightman/pytorch-image-models</a></p>\n\n<p>As the author of those packages I contribute with a permissive license so anyone can use them in applications like this. But as the author I would also like to have other users come back to the original source -- so that they have access to the latest versions, can provide feedback, bug reports, contribute via PR, etc. For a similar reason, a link to the original source in your kernels would be much appreciated. It is in fact required by Apache license terms to include some form of attribution for a redistribution of the code in documentation or file. Thanks!</p>",
          "votes": 26,
          "replies": []
        },
        {
          "id": 862679,
          "author_name": "Alex Shonenkov",
          "author_url": "",
          "post_date": "2020-05-26T18:28:06.803000",
          "content": "<p><a href=\"/rwightman\">@rwightman</a> Sure! I have added links on your repositories. Thank you for your work! </p>",
          "votes": 2,
          "replies": []
        },
        {
          "id": 862805,
          "author_name": "RossWightman",
          "author_url": "",
          "post_date": "2020-05-26T20:52:59.453000",
          "content": "<p><a href=\"/shonenkov\">@shonenkov</a> thanks, glad to see it's working well.. I've recently forwarded your train kernel link to a few who've asked about training on alternate datasets with different classes and augmentation pipelines as it's a concise example. I'll link it in the README if that sounds good with you?</p>",
          "votes": 2,
          "replies": []
        },
        {
          "id": 862826,
          "author_name": "Alex Shonenkov",
          "author_url": "",
          "post_date": "2020-05-26T21:21:43.307000",
          "content": "<p><a href=\"/rwightman\">@rwightman</a> It is really good idea! I think Global Wheat Detection is good example for training with custom datasets using your pytorch effdet implementations. It can be very useful for other users. Thank you a lot!</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 923137,
      "author_name": "Pragnesh Barik",
      "author_url": "",
      "post_date": "2020-07-10T14:56:38.453000",
      "content": "<p>Hi I am new, This comment is Non-Sense. Just to one up my tier.</p>",
      "votes": -13,
      "replies": []
    },
    {
      "id": 958128,
      "author_name": "JayChakalasiya",
      "author_url": "",
      "post_date": "2020-08-04T19:42:09.433000",
      "content": "<p><strong>We Are using these sources:</strong></p>\n\n<p><a href=\"https://github.com/rwightman/efficientdet-pytorch\">https://github.com/rwightman/efficientdet-pytorch</a></p>\n\n<p><a href=\"https://www.kaggle.com/shonenkov/timm-efficientdet-pytorch\">https://www.kaggle.com/shonenkov/timm-efficientdet-pytorch</a></p>\n\n<p><a href=\"https://www.kaggle.com/shonenkov/weightedboxesfusion\">https://www.kaggle.com/shonenkov/weightedboxesfusion</a></p>\n\n<p><a href=\"https://www.kaggle.com/hawkey/yolov5train\">https://www.kaggle.com/hawkey/yolov5train</a> \nOriginal Source:  <a href=\"https://github.com/ultralytics/yolov5\">https://github.com/ultralytics/yolov5</a></p>\n\n<p><strong>Configuration Files</strong>\n<a href=\"https://www.kaggle.com/orkatz2/configyolo5\">https://www.kaggle.com/orkatz2/configyolo5</a>\n<a href=\"https://www.kaggle.com/mihirmavalankar/omegaconf\">https://www.kaggle.com/mihirmavalankar/omegaconf</a>\n<a href=\"https://www.kaggle.com/shonenkov/omegaconf\">https://www.kaggle.com/shonenkov/omegaconf</a></p>\n\n<p><strong>Weights</strong>\n<a href=\"https://www.kaggle.com/jaychakalsiya/global-wheat-frcnn-23-06\">https://www.kaggle.com/jaychakalsiya/global-wheat-frcnn-23-06</a>\n<a href=\"https://www.kaggle.com/mihirmavalankar/detrmodels\">https://www.kaggle.com/mihirmavalankar/detrmodels</a>\n<a href=\"https://www.kaggle.com/mihirmavalankar/retinanetmodels\">https://www.kaggle.com/mihirmavalankar/retinanetmodels</a></p>\n\n<p>** Our Repo**\n<a href=\"https://github.com/jay-chakalasiya/Global-Wheat-Detction-Kaggle-2020.git\">https://github.com/jay-chakalasiya/Global-Wheat-Detction-Kaggle-2020.git</a>\nSources:\n<a href=\"https://github.com/zylo117/Yet-Another-EfficientDet-Pytorch\">https://github.com/zylo117/Yet-Another-EfficientDet-Pytorch</a>\n<a href=\"https://github.com/facebookresearch/detr\">https://github.com/facebookresearch/detr</a>\n<a href=\"https://github.com/yhenon/pytorch-retinanet\">https://github.com/yhenon/pytorch-retinanet</a>\n<a href=\"https://github.com/ultralytics/yolov5\">https://github.com/ultralytics/yolov5</a>\n<a href=\"https://github.com/ZFTurbo/Weighted-Boxes-Fusion\">https://github.com/ZFTurbo/Weighted-Boxes-Fusion</a></p>\n\n<p><em>Please comment if I'm forgetting to mention any of the original Git references,\nThank You</em></p>",
      "votes": -1,
      "replies": [
        {
          "id": 958133,
          "author_name": "Matt",
          "author_url": "",
          "post_date": "2020-08-04T19:46:39.340000",
          "content": "<p>Just so you know, reporting external data is due before the entry deadline, not the final deadline. Luckily it doesn't look like you're using anything that hasn't been posted here already :)</p>",
          "votes": 2,
          "replies": []
        }
      ]
    },
    {
      "id": 958904,
      "author_name": "Arseniy",
      "author_url": "",
      "post_date": "2020-08-05T07:57:06.390000",
      "content": "<p><a href=\"https://www.kaggle.com/shonenkov/inference-efficientdet\">https://www.kaggle.com/shonenkov/inference-efficientdet</a></p>\n\n<p><a href=\"https://www.kaggle.com/linyangsdu/detectors-wheat-inference-public\">https://www.kaggle.com/linyangsdu/detectors-wheat-inference-public</a></p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 950107,
      "author_name": "Shahrukh Khan",
      "author_url": "",
      "post_date": "2020-07-29T07:27:08.223000",
      "content": "<p>I am using/ experimenting with these pytorch pretrained weights as backbone, \n<code>\nmodel_urls = {\n    'fasterrcnn_resnet50_fpn_coco': 'https://download.pytorch.org/models/fasterrcnn_resnet50_fpn_coco-258fb6c6.pth',\n     'resnet18': 'https://download.pytorch.org/models/resnet18-5c106cde.pth',\n    'resnet34': 'https://download.pytorch.org/models/resnet34-333f7ec4.pth',\n    'resnet50': 'https://download.pytorch.org/models/resnet50-19c8e357.pth',\n    'resnet101': 'https://download.pytorch.org/models/resnet101-5d3b4d8f.pth',\n    'resnet152': 'https://download.pytorch.org/models/resnet152-b121ed2d.pth',\n    'resnext50_32x4d': 'https://download.pytorch.org/models/resnext50_32x4d-7cdf4587.pth',\n    'resnext101_32x8d': 'https://download.pytorch.org/models/resnext101_32x8d-8ba56ff5.pth',\n    'wide_resnet50_2': 'https://download.pytorch.org/models/wide_resnet50_2-95faca4d.pth',\n    'wide_resnet101_2': 'https://download.pytorch.org/models/wide_resnet101_2-32ee1156.pth',\n}\n</code></p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 949711,
      "author_name": "Ramadan",
      "author_url": "",
      "post_date": "2020-07-28T20:41:39.450000",
      "content": "<p><a href=\"https://www.kaggle.com/angqx95/effdet-pytorch\">https://www.kaggle.com/angqx95/effdet-pytorch</a>\n<a href=\"https://www.kaggle.com/angqx95/efficientdet-model\">https://www.kaggle.com/angqx95/efficientdet-model</a>\n<a href=\"https://www.kaggle.com/shonenkov/omegaconf\">https://www.kaggle.com/shonenkov/omegaconf</a>\n<a href=\"https://www.kaggle.com/kaushal2896/pycocotools\">https://www.kaggle.com/kaushal2896/pycocotools</a>\n<a href=\"https://www.kaggle.com/angqx95/torch-img-model\">https://www.kaggle.com/angqx95/torch-img-model</a>\n<a href=\"https://www.kaggle.com/shonenkov/weightedboxesfusion\">https://www.kaggle.com/shonenkov/weightedboxesfusion</a></p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 949351,
      "author_name": "team007",
      "author_url": "",
      "post_date": "2020-07-28T15:24:04.703000",
      "content": "<p>Tried some of these below, but no results so far.</p>\n\n<p><strong>Few keras/tensorflow references for object detection :</strong>\n<a href=\"https://github.com/google/automl/tree/master/efficientdet\">https://github.com/google/automl/tree/master/efficientdet</a>\n<a href=\"https://github.com/fizyr/keras-retinanet\">https://github.com/fizyr/keras-retinanet</a> and <a href=\"https://github.com/fizyr/tf-retinanet\">https://github.com/fizyr/tf-retinanet</a>\n<a href=\"https://github.com/xuannianz/EfficientDet\">https://github.com/xuannianz/EfficientDet</a>\n<a href=\"https://github.com/calmisential/EfficientDet_TensorFlow2\">https://github.com/calmisential/EfficientDet_TensorFlow2</a></p>\n\n<p>*<em>Pre trained Model files : *</em>\n<a href=\"https://github.com/tensorflow/\">https://github.com/tensorflow/</a>\n<a href=\"https://github.com/qubvel/efficientnet\">https://github.com/qubvel/efficientnet</a>\n<a href=\"https://github.com/Callidior/keras-applications/releases\">https://github.com/Callidior/keras-applications/releases</a></p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 948451,
      "author_name": "XY",
      "author_url": "",
      "post_date": "2020-07-28T00:51:00.690000",
      "content": "<p>Thanks <a href=\"/whurobin\">@whurobin</a> \n<a href=\"https://github.com/wuxinwang1997/wheatdetection\">https://github.com/wuxinwang1997/wheatdetection</a>\n<a href=\"https://www.kaggle.com/whurobin/wbf-over-tta-single-model-fasterrcnn-resnest\">https://www.kaggle.com/whurobin/wbf-over-tta-single-model-fasterrcnn-resnest</a></p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 947040,
      "author_name": "outrunner",
      "author_url": "",
      "post_date": "2020-07-27T03:22:14.920000",
      "content": "<p>keras-centernet\n<a href=\"https://github.com/see--/keras-centernet\">https://github.com/see--/keras-centernet</a></p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 946410,
      "author_name": "Costas Voglis",
      "author_url": "",
      "post_date": "2020-07-26T15:01:16.553000",
      "content": "<p>Spike wheat dataset <a href=\"https://www.kaggle.com/voglinio/spike-wheat-dataset\">https://www.kaggle.com/voglinio/spike-wheat-dataset</a></p>",
      "votes": 0,
      "replies": [
        {
          "id": 946709,
          "author_name": "Matt",
          "author_url": "",
          "post_date": "2020-07-26T19:02:07.107000",
          "content": "<p>I'm curious if you saw any improvement with it? I noticed the labels are pretty noisy and the photos are taken from a side view instead of top-down</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 947828,
          "author_name": "tereka",
          "author_url": "",
          "post_date": "2020-07-27T14:00:21.690000",
          "content": "<p>Can we use this dataset? no license is similar MIT license?(I don't know license in details.)\n<a href=\"https://www.kaggle.com/c/global-wheat-detection/discussion/164346#917316\">https://www.kaggle.com/c/global-wheat-detection/discussion/164346#917316</a></p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 947989,
          "author_name": "Costas Voglis",
          "author_url": "",
          "post_date": "2020-07-27T15:38:36.313000",
          "content": "<p>Not that much of improvement.. and since it is a little fuzzy whether we could use it I didn't push it that much. </p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 946200,
      "author_name": "Zikai",
      "author_url": "",
      "post_date": "2020-07-26T12:34:21.223000",
      "content": "<p>PP-YOLO with PaddleDetection\n<a href=\"https://github.com/PaddlePaddle/PaddleDetection\">https://github.com/PaddlePaddle/PaddleDetection</a>\n<a href=\"https://arxiv.org/abs/2007.12099\">https://arxiv.org/abs/2007.12099</a></p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 943338,
      "author_name": "DrP",
      "author_url": "",
      "post_date": "2020-07-24T09:54:38.833000",
      "content": "<p>External data:\n<a href=\"https://www.kaggle.com/ravirajsinh45/crop-and-weed-detection-data-with-bounding-boxes\">https://www.kaggle.com/ravirajsinh45/crop-and-weed-detection-data-with-bounding-boxes</a>?</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 941434,
      "author_name": "nvnn",
      "author_url": "",
      "post_date": "2020-07-23T08:07:40.673000",
      "content": "<p>torchvision pretrained models: <a href=\"https://pytorch.org/docs/stable/torchvision/models.html\">https://pytorch.org/docs/stable/torchvision/models.html</a>\nCOCO dataset: <a href=\"https://cocodataset.org/\">https://cocodataset.org/</a></p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 937434,
      "author_name": "Yifan Wang",
      "author_url": "",
      "post_date": "2020-07-21T02:49:04.020000",
      "content": "<p><a href=\"https://github.com/ZFTurbo/Weighted-Boxes-Fusion/tree/v1.0.5\" target=\"_blank\">https://github.com/ZFTurbo/Weighted-Boxes-Fusion/tree/v1.0.5</a><br>\n<a href=\"https://github.com/NVIDIA/apex\" target=\"_blank\">https://github.com/NVIDIA/apex</a><br>\n<a href=\"https://github.com/ultralytics/yolov5\" target=\"_blank\">https://github.com/ultralytics/yolov5</a></p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 937195,
      "author_name": "Ryan Tran",
      "author_url": "",
      "post_date": "2020-07-20T19:45:23.940000",
      "content": "<p><a href=\"https://github.com/rwightman/efficientdet-pytorch\">https://github.com/rwightman/efficientdet-pytorch</a>\n<a href=\"https://github.com/rwightman/pytorch-image-models\">https://github.com/rwightman/pytorch-image-models</a>\n<a href=\"https://github.com/ZFTurbo/Weighted-Boxes-Fusion\">https://github.com/ZFTurbo/Weighted-Boxes-Fusion</a>\n<a href=\"https://www.kaggle.com/greatgamedota/timm-package\">https://www.kaggle.com/greatgamedota/timm-package</a>\n<a href=\"https://www.kaggle.com/kaushal2896/pycocotools\">https://www.kaggle.com/kaushal2896/pycocotools</a>\n<a href=\"https://www.kaggle.com/shonenkov/omegaconf\">https://www.kaggle.com/shonenkov/omegaconf</a>\n<a href=\"https://www.kaggle.com/shonenkov/weightedboxesfusion\">https://www.kaggle.com/shonenkov/weightedboxesfusion</a>\n<a href=\"https://www.kaggle.com/shonenkov/timm-efficientdet-pytorch\">https://www.kaggle.com/shonenkov/timm-efficientdet-pytorch</a>\n<a href=\"https://www.kaggle.com/mathurinache/efficientdet\">https://www.kaggle.com/mathurinache/efficientdet</a></p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 936666,
      "author_name": "alexx",
      "author_url": "",
      "post_date": "2020-07-20T12:30:20.423000",
      "content": "<p><a href=\"https://github.com/NVIDIA/apex\" target=\"_blank\">https://github.com/NVIDIA/apex</a><br>\n<a href=\"https://www.kaggle.com/ii5m0k3ii/nvidia-apex\" target=\"_blank\">https://www.kaggle.com/ii5m0k3ii/nvidia-apex</a></p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 936134,
      "author_name": "李先明",
      "author_url": "",
      "post_date": "2020-07-20T02:36:34.987000",
      "content": "<p><a href=\"https://github.com/rwightman/efficientdet-pytorch\">https://github.com/rwightman/efficientdet-pytorch</a>\n<a href=\"https://github.com/rwightman/pytorch-image-models\">https://github.com/rwightman/pytorch-image-models</a>\n<a href=\"https://www.kaggle.com/greatgamedota/timm-package\">https://www.kaggle.com/greatgamedota/timm-package</a>\n<a href=\"https://www.kaggle.com/kaushal2896/pycocotools\">https://www.kaggle.com/kaushal2896/pycocotools</a>\n<a href=\"https://www.kaggle.com/shonenkov/omegaconf\">https://www.kaggle.com/shonenkov/omegaconf</a>\n<a href=\"https://www.kaggle.com/mathurinache/efficientdet\">https://www.kaggle.com/mathurinache/efficientdet</a></p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 935701,
      "author_name": "Mark Peng",
      "author_url": "",
      "post_date": "2020-07-19T15:15:12.213000",
      "content": "<p>Spike Dataset:<br>\n<a href=\"https://sourceforge.net/projects/spike-dataset/\" target=\"_blank\">https://sourceforge.net/projects/spike-dataset/</a></p>\n<p>YOLOv4:</p>\n<ul>\n<li><a href=\"https://github.com/AlexeyAB/darknet\" target=\"_blank\">https://github.com/AlexeyAB/darknet</a></li>\n</ul>\n<p>EfficientDet:</p>\n<ul>\n<li><a href=\"https://github.com/rwightman/efficientdet-pytorch\" target=\"_blank\">https://github.com/rwightman/efficientdet-pytorch</a></li>\n<li><a href=\"https://github.com/rwightman/pytorch-image-models\" target=\"_blank\">https://github.com/rwightman/pytorch-image-models</a></li>\n<li><a href=\"https://github.com/google/automl/tree/master/efficientdet\" target=\"_blank\">https://github.com/google/automl/tree/master/efficientdet</a></li>\n<li><a href=\"https://github.com/zylo117/Yet-Another-EfficientDet-Pytorch\" target=\"_blank\">https://github.com/zylo117/Yet-Another-EfficientDet-Pytorch</a></li>\n</ul>\n<p>RetinaNet:</p>\n<ul>\n<li><a href=\"https://github.com/yhenon/pytorch-retinanet\" target=\"_blank\">https://github.com/yhenon/pytorch-retinanet</a></li>\n</ul>\n<p>CenterNet:</p>\n<ul>\n<li><a href=\"https://github.com/xingyizhou/CenterNet\" target=\"_blank\">https://github.com/xingyizhou/CenterNet</a></li>\n</ul>\n<p>ResNeSt:</p>\n<ul>\n<li><a href=\"https://github.com/zhanghang1989/detectron2-ResNeSt\" target=\"_blank\">https://github.com/zhanghang1989/detectron2-ResNeSt</a></li>\n</ul>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 931291,
      "author_name": "arutema47",
      "author_url": "",
      "post_date": "2020-07-16T05:29:21.170000",
      "content": "<p>Can we use youtube videos to weakly supervise our models?\ne.g. facebook uses Instagram photos to get better performance.</p>\n\n<p>Maybe this isn't allowed because of the very strict license issues but seems like a interesting approach (for CVPRR).</p>\n\n<p>some wheat videos\n<a href=\"https://www.youtube.com/watch?v=YTix0yymVWM\">Agriculture Expert Guide to Increase Wheat Production. Wheat production in Pakistan</a></p>",
      "votes": 0,
      "replies": [
        {
          "id": 931303,
          "author_name": "Matt",
          "author_url": "",
          "post_date": "2020-07-16T05:36:28.593000",
          "content": "<p>Not sure what you mean exactly. The video you shared (and many on the same channel) does not contain any video of wheat itself, just people discussing wheat agriculture.</p>\n\n<p>Personally, I would refrain from using any youtube content for the competition. Even if it is licensed properly there have been big issues in the past with this (<a href=\"https://www.kaggle.com/c/deepfake-detection-challenge/discussion/157983\">source</a>).</p>\n\n<p>But if you are able to come up with something for CVPRR, I think it's a great idea!</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 921474,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-07-09T10:20:02.417000",
      "content": "",
      "votes": 0,
      "replies": []
    },
    {
      "id": 918083,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-07-07T01:07:42.593000",
      "content": "",
      "votes": 0,
      "replies": [
        {
          "id": 918106,
          "author_name": "",
          "author_url": "",
          "post_date": "2020-07-07T01:55:01.180000",
          "content": "",
          "votes": 2,
          "replies": []
        }
      ]
    },
    {
      "id": 916803,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-07-06T03:46:55.407000",
      "content": "",
      "votes": 0,
      "replies": [
        {
          "id": 916852,
          "author_name": "",
          "author_url": "",
          "post_date": "2020-07-06T04:46:01",
          "content": "",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 916274,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-07-05T14:21:59.270000",
      "content": "",
      "votes": 0,
      "replies": [
        {
          "id": 917540,
          "author_name": "",
          "author_url": "",
          "post_date": "2020-07-06T15:45:47.073000",
          "content": "",
          "votes": 0,
          "replies": []
        },
        {
          "id": 917850,
          "author_name": "",
          "author_url": "",
          "post_date": "2020-07-06T19:10:44.170000",
          "content": "",
          "votes": 0,
          "replies": []
        },
        {
          "id": 919750,
          "author_name": "",
          "author_url": "",
          "post_date": "2020-07-08T04:57:58.713000",
          "content": "",
          "votes": -2,
          "replies": []
        },
        {
          "id": 919794,
          "author_name": "",
          "author_url": "",
          "post_date": "2020-07-08T05:36:05.633000",
          "content": "",
          "votes": 1,
          "replies": []
        },
        {
          "id": 920584,
          "author_name": "",
          "author_url": "",
          "post_date": "2020-07-08T17:35:14.077000",
          "content": "",
          "votes": 1,
          "replies": []
        },
        {
          "id": 927017,
          "author_name": "",
          "author_url": "",
          "post_date": "2020-07-13T06:41:43.603000",
          "content": "",
          "votes": 0,
          "replies": []
        },
        {
          "id": 927075,
          "author_name": "",
          "author_url": "",
          "post_date": "2020-07-13T07:24:42.460000",
          "content": "",
          "votes": 0,
          "replies": []
        },
        {
          "id": 929309,
          "author_name": "",
          "author_url": "",
          "post_date": "2020-07-14T15:45:41.537000",
          "content": "",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 915096,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-07-04T13:19:47.593000",
      "content": "",
      "votes": 0,
      "replies": []
    },
    {
      "id": 914818,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-07-04T09:00:24.647000",
      "content": "",
      "votes": 0,
      "replies": []
    },
    {
      "id": 909935,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-07-01T00:08:41.930000",
      "content": "",
      "votes": 0,
      "replies": []
    },
    {
      "id": 901619,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-06-25T15:51:47.467000",
      "content": "",
      "votes": 0,
      "replies": []
    },
    {
      "id": 899121,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-06-24T02:12:21.780000",
      "content": "",
      "votes": 0,
      "replies": []
    },
    {
      "id": 899068,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-06-24T00:20:42.397000",
      "content": "",
      "votes": 0,
      "replies": []
    },
    {
      "id": 899062,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-06-24T00:09:16.540000",
      "content": "",
      "votes": 0,
      "replies": []
    },
    {
      "id": 899054,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-06-23T23:55:47.407000",
      "content": "",
      "votes": 0,
      "replies": []
    },
    {
      "id": 881871,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-06-11T12:30:54.147000",
      "content": "",
      "votes": 0,
      "replies": [
        {
          "id": 882073,
          "author_name": "",
          "author_url": "",
          "post_date": "2020-06-11T15:08:36.283000",
          "content": "",
          "votes": 0,
          "replies": []
        },
        {
          "id": 882123,
          "author_name": "",
          "author_url": "",
          "post_date": "2020-06-11T15:45:45.563000",
          "content": "",
          "votes": 0,
          "replies": []
        },
        {
          "id": 882190,
          "author_name": "",
          "author_url": "",
          "post_date": "2020-06-11T16:30:28.843000",
          "content": "",
          "votes": 0,
          "replies": []
        },
        {
          "id": 882215,
          "author_name": "",
          "author_url": "",
          "post_date": "2020-06-11T16:45:00.350000",
          "content": "",
          "votes": 0,
          "replies": []
        },
        {
          "id": 882275,
          "author_name": "",
          "author_url": "",
          "post_date": "2020-06-11T17:16:34.113000",
          "content": "",
          "votes": 2,
          "replies": []
        },
        {
          "id": 882315,
          "author_name": "",
          "author_url": "",
          "post_date": "2020-06-11T18:00:45.730000",
          "content": "",
          "votes": 2,
          "replies": []
        },
        {
          "id": 882505,
          "author_name": "",
          "author_url": "",
          "post_date": "2020-06-11T21:29:03.703000",
          "content": "",
          "votes": 0,
          "replies": []
        },
        {
          "id": 888650,
          "author_name": "",
          "author_url": "",
          "post_date": "2020-06-16T13:36:45.733000",
          "content": "",
          "votes": 2,
          "replies": []
        },
        {
          "id": 899327,
          "author_name": "",
          "author_url": "",
          "post_date": "2020-06-24T07:00:25.563000",
          "content": "",
          "votes": 2,
          "replies": []
        }
      ]
    },
    {
      "id": 863159,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-05-27T05:37:19.287000",
      "content": "",
      "votes": 0,
      "replies": [
        {
          "id": 863222,
          "author_name": "",
          "author_url": "",
          "post_date": "2020-05-27T06:51:30.907000",
          "content": "",
          "votes": 0,
          "replies": []
        },
        {
          "id": 864294,
          "author_name": "",
          "author_url": "",
          "post_date": "2020-05-28T00:01:16.810000",
          "content": "",
          "votes": 0,
          "replies": []
        },
        {
          "id": 879104,
          "author_name": "",
          "author_url": "",
          "post_date": "2020-06-09T07:36:49.307000",
          "content": "",
          "votes": -1,
          "replies": []
        }
      ]
    },
    {
      "id": 859493,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-05-24T13:37:08.300000",
      "content": "",
      "votes": 0,
      "replies": []
    },
    {
      "id": 851855,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-05-18T01:05:33.437000",
      "content": "",
      "votes": 0,
      "replies": []
    },
    {
      "id": 850092,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-05-16T10:36:06.560000",
      "content": "",
      "votes": 0,
      "replies": [
        {
          "id": 919792,
          "author_name": "",
          "author_url": "",
          "post_date": "2020-07-08T05:35:36.877000",
          "content": "",
          "votes": 0,
          "replies": []
        },
        {
          "id": 942745,
          "author_name": "",
          "author_url": "",
          "post_date": "2020-07-24T01:23:46.313000",
          "content": "",
          "votes": -2,
          "replies": []
        }
      ]
    },
    {
      "id": 834373,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-05-05T14:12:20.493000",
      "content": "",
      "votes": 0,
      "replies": [
        {
          "id": 845596,
          "author_name": "",
          "author_url": "",
          "post_date": "2020-05-13T10:02:47.830000",
          "content": "",
          "votes": 0,
          "replies": []
        },
        {
          "id": 853692,
          "author_name": "",
          "author_url": "",
          "post_date": "2020-05-19T12:06:49.457000",
          "content": "",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 927147,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-07-13T08:21:48.627000",
      "content": "",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "947309": "some nice and clear pictures we took in our hometown Wuhan and in another beautiful city Paris, feel free to use them.\nhttps://www.kaggle.com/meliao/no-wheat-image",
    "859526": "@maggiemd I'm a little confused as to what constitutes \"external data sources\" here. I notice other Kagglers posting model weights and python packages. Do these have to be reported? (I've checked the competition rules and it wasn't clear to me) Thank you",
    "833427": "Per the competition rules, post links to your external data sources here before the deadline specified. Once it has been posted, you do not need to post it again.",
    "836376": "I'm looking to use an external dataset to expand the number of negative samples. The best thing I've found so far is\nhttps://www.kaggle.com/ravirajsinh45/crop-and-weed-detection-data-with-bounding-boxes#agri_0_1017.txt\nthough I'm not too confident this will actually be of any use. If anyone else has datasets in mind for this task, I'd appreciate hearing about it.",
    "955649": "@maggiemd Are optimizers and schedulers part of the External Data Thread and need to be reported? Have not seen that kind of post before in any other competition, like for AdamW,Radam,schedulers,dataloaders that are not built-in, so it's good to know, even if it's to late this time to post/report.",
    "951191": "Hello,If i use the dataset someone public in Kaggle,should I post it?",
    "950446": "I may be using these\n\nhttps://ai.googleblog.com/2020/05/open-sourcing-bit-exploring-large-scale.html\nModels from PyTorch hub",
    "950265": "Hello,can I process the train.csv and use this instead of the original one in the notebook I submitted?Thanks",
    "948237": "[Wheat Ear Dataset ](https://zenodo.org/record/1471626#.W9H_XFUzZjU)",
    "872875": "The pre-trained models from GluonCV:\n\n* [https://gluon-cv.mxnet.io/model_zoo/detection.html](https://gluon-cv.mxnet.io/model_zoo/detection.html)\n\nWeighted-Boxes-Fusion:\n\n* [https://github.com/ZFTurbo/Weighted-Boxes-Fusion](https://github.com/ZFTurbo/Weighted-Boxes-Fusion)",
    "949765": "\nhttps://github.com/open-mmlab/mmdetection\nhttps://github.com/VDIGPKU/CBNet\nhttps://github.com/JialeCao001/D2Det\nhttps://github.com/shinya7y/UniverseNet\nhttps://github.com/thangvubk/Cascade-RPN\nhttps://github.com/WXinlong/SOLO\nhttps://github.com/Sense-X/TSD\nhttps://github.com/rwightman/efficientdet-pytorch\nhttps://github.com/rwightman/pytorch-image-models\nhttps://github.com/zylo117/Yet-Another-EfficientDet-Pytorch\nhttps://github.com/joe-siyuan-qiao/DetectoRS\nhttps://github.com/facebookresearch/detectron2\nhttps://github.com/TuSimple/simpledet\nhttps://github.com/naoto0804/pytorch-AdaIN\nhttps://github.com/eriklindernoren/PyTorch-GAN\nhttps://github.com/NVlabs/stylegan2\nhttps://github.com/clovaai/stargan-v2\nhttps://github.com/junyanz/pytorch-CycleGAN-and-pix2pix",
    "853968": "https://github.com/rwightman/efficientdet-pytorch\nhttps://github.com/rwightman/pytorch-image-models\nhttps://www.kaggle.com/greatgamedota/timm-package\nhttps://www.kaggle.com/kaushal2896/pycocotools\nhttps://www.kaggle.com/shonenkov/omegaconf\nhttps://www.kaggle.com/mathurinache/efficientdet",
    "923137": "Hi I am new, This comment is Non-Sense. Just to one up my tier.\n",
    "958128": "**We Are using these sources:**\n\nhttps://github.com/rwightman/efficientdet-pytorch\n\nhttps://www.kaggle.com/shonenkov/timm-efficientdet-pytorch\n\nhttps://www.kaggle.com/shonenkov/weightedboxesfusion\n\nhttps://www.kaggle.com/hawkey/yolov5train \nOriginal Source:  https://github.com/ultralytics/yolov5\n\n**Configuration Files**\nhttps://www.kaggle.com/orkatz2/configyolo5\nhttps://www.kaggle.com/mihirmavalankar/omegaconf\nhttps://www.kaggle.com/shonenkov/omegaconf\n\n**Weights**\nhttps://www.kaggle.com/jaychakalsiya/global-wheat-frcnn-23-06\nhttps://www.kaggle.com/mihirmavalankar/detrmodels\nhttps://www.kaggle.com/mihirmavalankar/retinanetmodels\n\n** Our Repo**\nhttps://github.com/jay-chakalasiya/Global-Wheat-Detction-Kaggle-2020.git\nSources:\nhttps://github.com/zylo117/Yet-Another-EfficientDet-Pytorch\nhttps://github.com/facebookresearch/detr\nhttps://github.com/yhenon/pytorch-retinanet\nhttps://github.com/ultralytics/yolov5\nhttps://github.com/ZFTurbo/Weighted-Boxes-Fusion\n\n\n*Please comment if I'm forgetting to mention any of the original Git references,\nThank You*\n",
    "958904": "https://www.kaggle.com/shonenkov/inference-efficientdet\n\nhttps://www.kaggle.com/linyangsdu/detectors-wheat-inference-public",
    "950107": "I am using/ experimenting with these pytorch pretrained weights as backbone, \n```\nmodel_urls = {\n    'fasterrcnn_resnet50_fpn_coco': 'https://download.pytorch.org/models/fasterrcnn_resnet50_fpn_coco-258fb6c6.pth',\n     'resnet18': 'https://download.pytorch.org/models/resnet18-5c106cde.pth',\n    'resnet34': 'https://download.pytorch.org/models/resnet34-333f7ec4.pth',\n    'resnet50': 'https://download.pytorch.org/models/resnet50-19c8e357.pth',\n    'resnet101': 'https://download.pytorch.org/models/resnet101-5d3b4d8f.pth',\n    'resnet152': 'https://download.pytorch.org/models/resnet152-b121ed2d.pth',\n    'resnext50_32x4d': 'https://download.pytorch.org/models/resnext50_32x4d-7cdf4587.pth',\n    'resnext101_32x8d': 'https://download.pytorch.org/models/resnext101_32x8d-8ba56ff5.pth',\n    'wide_resnet50_2': 'https://download.pytorch.org/models/wide_resnet50_2-95faca4d.pth',\n    'wide_resnet101_2': 'https://download.pytorch.org/models/wide_resnet101_2-32ee1156.pth',\n}\n```",
    "949711": "https://www.kaggle.com/angqx95/effdet-pytorch\nhttps://www.kaggle.com/angqx95/efficientdet-model\nhttps://www.kaggle.com/shonenkov/omegaconf\nhttps://www.kaggle.com/kaushal2896/pycocotools\nhttps://www.kaggle.com/angqx95/torch-img-model\nhttps://www.kaggle.com/shonenkov/weightedboxesfusion",
    "949351": "Tried some of these below, but no results so far.\n\n**Few keras/tensorflow references for object detection :**\nhttps://github.com/google/automl/tree/master/efficientdet\nhttps://github.com/fizyr/keras-retinanet and https://github.com/fizyr/tf-retinanet\nhttps://github.com/xuannianz/EfficientDet\nhttps://github.com/calmisential/EfficientDet_TensorFlow2\n\n**Pre trained Model files : **\nhttps://github.com/tensorflow/\nhttps://github.com/qubvel/efficientnet\nhttps://github.com/Callidior/keras-applications/releases\n\n",
    "948451": "Thanks @whurobin \nhttps://github.com/wuxinwang1997/wheatdetection\nhttps://www.kaggle.com/whurobin/wbf-over-tta-single-model-fasterrcnn-resnest",
    "947040": "keras-centernet\n[https://github.com/see--/keras-centernet](https://github.com/see--/keras-centernet)",
    "946410": "Spike wheat dataset https://www.kaggle.com/voglinio/spike-wheat-dataset",
    "946200": "PP-YOLO with PaddleDetection\nhttps://github.com/PaddlePaddle/PaddleDetection\nhttps://arxiv.org/abs/2007.12099",
    "943338": "External data:\nhttps://www.kaggle.com/ravirajsinh45/crop-and-weed-detection-data-with-bounding-boxes?",
    "941434": "torchvision pretrained models: https://pytorch.org/docs/stable/torchvision/models.html\nCOCO dataset: https://cocodataset.org/",
    "937434": "https://github.com/ZFTurbo/Weighted-Boxes-Fusion/tree/v1.0.5\nhttps://github.com/NVIDIA/apex\nhttps://github.com/ultralytics/yolov5\n",
    "937195": "https://github.com/rwightman/efficientdet-pytorch\nhttps://github.com/rwightman/pytorch-image-models\nhttps://github.com/ZFTurbo/Weighted-Boxes-Fusion\nhttps://www.kaggle.com/greatgamedota/timm-package\nhttps://www.kaggle.com/kaushal2896/pycocotools\nhttps://www.kaggle.com/shonenkov/omegaconf\nhttps://www.kaggle.com/shonenkov/weightedboxesfusion\nhttps://www.kaggle.com/shonenkov/timm-efficientdet-pytorch\nhttps://www.kaggle.com/mathurinache/efficientdet",
    "936666": "https://github.com/NVIDIA/apex\nhttps://www.kaggle.com/ii5m0k3ii/nvidia-apex",
    "936134": "https://github.com/rwightman/efficientdet-pytorch\nhttps://github.com/rwightman/pytorch-image-models\nhttps://www.kaggle.com/greatgamedota/timm-package\nhttps://www.kaggle.com/kaushal2896/pycocotools\nhttps://www.kaggle.com/shonenkov/omegaconf\nhttps://www.kaggle.com/mathurinache/efficientdet",
    "935701": "Spike Dataset:\nhttps://sourceforge.net/projects/spike-dataset/\n\nYOLOv4:\n* https://github.com/AlexeyAB/darknet\n\nEfficientDet:\n* https://github.com/rwightman/efficientdet-pytorch\n* https://github.com/rwightman/pytorch-image-models\n* https://github.com/google/automl/tree/master/efficientdet\n* https://github.com/zylo117/Yet-Another-EfficientDet-Pytorch\n\nRetinaNet:\n* https://github.com/yhenon/pytorch-retinanet\n\nCenterNet:\n* https://github.com/xingyizhou/CenterNet\n\nResNeSt:\n* https://github.com/zhanghang1989/detectron2-ResNeSt",
    "931291": "Can we use youtube videos to weakly supervise our models?\ne.g. facebook uses Instagram photos to get better performance.\n\nMaybe this isn't allowed because of the very strict license issues but seems like a interesting approach (for CVPRR).\n\nsome wheat videos\n[Agriculture Expert Guide to Increase Wheat Production. Wheat production in Pakistan](https://www.youtube.com/watch?v=YTix0yymVWM)\n\n",
    "921474": "I know that yolov5 has issues.\nAnd I'm not sure how to share it, but I leave a post in this thread.\n\n[https://github.com/ultralytics/yolov5](https://github.com/ultralytics/yolov5)\n[https://www.kaggle.com/shonenkov/weightedboxesfusion](https://www.kaggle.com/shonenkov/weightedboxesfusion)\n[https://github.com/rwightman/efficientdet-pytorch](https://github.com/rwightman/efficientdet-pytorch)\n[https://github.com/rwightman/pytorch-image-models](https://github.com/rwightman/pytorch-image-models)\n[https://github.com/zylo117/Yet-Another-EfficientDet-Pytorch](https://github.com/zylo117/Yet-Another-EfficientDet-Pytorch)\n[https://github.com/open-mmlab/mmdetection](https://github.com/open-mmlab/mmdetection)\n[https://github.com/facebookresearch/detr/](https://github.com/facebookresearch/detr/)",
    "918083": "\"Once it has been posted, you do not need to post it again.\" Does that mean if someone posted YOLOv5 everyone doesn't have to post it? I'm asking because I'm seeing people posting the same things over and over. I just want to be clear if we have to post our own external data or ?? ",
    "916803": "Pretrained models using 400K synthetic data: \nhttps://csiro-robotics.github.io/UPGen-Webpage/\n\n",
    "916274": "What about Synthetic data? Is that considered as external data?",
    "915096": "model -\nhttps://github.com/ultralytics/yolov5\nWBM-\nhttps://www.kaggle.com/shonenkov/weightedboxesfusion",
    "914818": "model -\nhttps://github.com/ultralytics/yolov5\nWBM-\nhttps://www.kaggle.com/shonenkov/weightedboxesfusion",
    "909935": "RESNEXT WSL\nhttps://pytorch.org/hub/facebookresearch_WSL-Images_resnext/",
    "901619": "https://www.kaggle.com/imazekishota/yolov5-output-conf",
    "899121": "https://github.com/tensorflow/models",
    "899068": "https://www.kaggle.com/orkatz2/yolov5-train",
    "899062": "Weighted boxes fusion:\nhttps://www.kaggle.com/shonenkov/weightedboxesfusion",
    "899054": "YOLOv5:\nhttps://github.com/ultralytics/yolov5",
    "881871": "@maggiemd Quick question about external data - I see that we are not allowed to use custom packages, but that we are allowed to use external data. Does external data include open-source .py class files? \n\nIf we can upload .py files, at what point does it start to be considered a 'custom package'; as long as it's not compiled it's fine?",
    "863159": "https://plantimages.nottingham.ac.uk/\nhttps://pytorch.org/docs/stable/torchvision/models.html",
    "859493": "https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix\n",
    "851855": "1. https://github.com/zylo117/Yet-Another-EfficientDet-Pytorch\n2. https://github.com/open-mmlab/mmdetection",
    "850092": "Out of curiosity, what if I use pictures I made myself ? Should they be uploaded somewhere ?",
    "834373": "Mask-RCNN (matterport) source code. \n- https://www.kaggle.com/ipythonx/maskrcnn-keras-source-code\n\nBackbone weights:\n- https://www.kaggle.com/ipythonx/keras-pretrained-imagenet-weights\n- https://www.kaggle.com/ipythonx/efficientnet-keras-noisystudent-weights-b0b7\n- https://www.kaggle.com/ipythonx/cocowg",
    "927147": ""
  }
}