Thanks everyone for your participation in the Global Wheat Detection Competition!
\n\nPlease note that the Private leaderboard will be rerun with the updated private test set, so please stay tuned for the finalized leaderboard within the next couple days.
\n\nI also wanted to remind everyone that the deadline for paper submissions to the Computer Vision Problems in Plant Phenotyping workshop is Aug. 9 on CMT (https://cmt3.research.microsoft.com/GWC2020). If you plan to submit a paper, please send a short email to ian.stavness@usask.ca to let us know.
\n\nFollowing discussion in the forums, we understand some participants have re-labeled the training datasets. If you are willing to share these annotations, please email ian.stavness@usask.ca This will be a big help for us to refine the annotations for future competitions.
", "messages": [ { "id": 962179, "postDate": "2020-08-07T22:01:30.697Z", "content": "We are working on the finalized leaderboard, but it will still take a few more days. Thanks everyone for your patience.
", "rawMarkdown": "We are working on the finalized leaderboard, but it will still take a few more days. Thanks everyone for your patience.", "votes": 20, "replies": [ { "id": 962506, "postDate": "2020-08-08T07:17:08.870Z", "content": "Will you remove all non-compliant submissions from leaderboard?
", "rawMarkdown": "Will you remove all non-compliant submissions from leaderboard?", "votes": 1 }, { "id": 962516, "postDate": "2020-08-08T07:29:05.513Z", "content": "It seems like they will. However, the question is how many places (or what part of the LB) will be checked. it's possible that after removing a lot of Yolov5 users from the medal zone this will fill up with the non-compliant submissions again so they shall check people in this zone again.
", "rawMarkdown": "It seems like they will. However, the question is how many places (or what part of the LB) will be checked. it's possible that after removing a lot of Yolov5 users from the medal zone this will fill up with the non-compliant submissions again so they shall check people in this zone again.", "votes": 1 } ] }, { "id": 959542, "postDate": "2020-08-05T17:17:36.430Z", "content": "Thanks Ian, I did not realize this was not the corrected test set.
\n\nCould you please answer if submissions will be reviewed for license compliance? This has been a big issue in this competition with YoloV5 and it would be unfair after all this if those who abided by the rules were stuck at these standings.
", "rawMarkdown": "Thanks Ian, I did not realize this was not the corrected test set.\n\nCould you please answer if submissions will be reviewed for license compliance? This has been a big issue in this competition with YoloV5 and it would be unfair after all this if those who abided by the rules were stuck at these standings. ", "votes": 11, "replies": [ { "id": 959662, "postDate": "2020-08-05T19:44:49.467Z", "content": "Yes, Kaggle has a standard procedure for finalizing the leaderboard and verifying compliance of winning submissions, which we are following.
", "rawMarkdown": "Yes, Kaggle has a standard procedure for finalizing the leaderboard and verifying compliance of winning submissions, which we are following.", "votes": -15 }, { "id": 959691, "postDate": "2020-08-05T20:26:39.420Z", "content": "Hello! What do you mean by winning solutions? Will you check only gold medal submits or the entire medal zone as well?
", "rawMarkdown": "Hello! What do you mean by winning solutions? Will you check only gold medal submits or the entire medal zone as well?", "votes": 14 }, { "id": 959775, "postDate": "2020-08-05T22:04:19.743Z", "content": "Verifying compliance of winning submissions only????? This is ridiculous
", "rawMarkdown": "Verifying compliance of **winning submissions** only????? This is ridiculous", "votes": 19 }, { "id": 959810, "postDate": "2020-08-05T23:04:25.100Z", "content": "The kaggle team worker has told us that we canβt get a medal by YOLO V5 (as the discussion mentioned ) resulting in most of us didn't choosed YOLO V5 as final submission.So i believe kaggle must can give us a fair result :D
", "rawMarkdown": "The kaggle team worker has told us that we canβt get a medal by YOLO V5 (as the [discussion](https://www.kaggle.com/c/global-wheat-detection/discussion/167140) mentioned ) resulting in most of us didn't choosed YOLO V5 as final submission.So i believe kaggle must can give us a fair result :D", "votes": 19 }, { "id": 959887, "postDate": "2020-08-06T02:08:19.297Z", "content": "If you check only a few kernels in the top tier, it is a deception for all participants who have followed the rules.
", "rawMarkdown": "If you check only a few kernels in the top tier, it is a deception for all participants who have followed the rules.", "votes": 11 }, { "id": 960089, "postDate": "2020-08-06T06:09:53.280Z", "content": "As many have asked what is really a \"winning solution\" olny gold region or the whole medal zone ??\nCan you shed some light\n@stavness \nIs this a fair with every one who obeyed the rules \nNeed some answers
", "rawMarkdown": "As many have asked what is really a \"winning solution\" olny gold region or the whole medal zone ??\nCan you shed some light\n@stavness \nIs this a fair with every one who obeyed the rules \nNeed some answers", "votes": 1 }, { "id": 960305, "postDate": "2020-08-06T09:47:08.507Z", "content": "i wish i would have selected yolo v5 then would have got medal. ILooks like Medal positions may remain stand still this is just for winning solution.
", "rawMarkdown": " \ni wish i would have selected yolo v5 then would have got medal. ILooks like Medal positions may remain stand still this is just for winning solution. \n", "votes": -1 }, { "id": 960323, "postDate": "2020-08-06T10:03:06.420Z", "content": "We trusted kaggle and submitted complient solution
", "rawMarkdown": "We trusted kaggle and submitted complient solution", "votes": 5 } ] }, { "id": 959923, "postDate": "2020-08-06T03:08:06.607Z", "content": "This is ridiculous. Because we train on hidden testset with augmentation, we get different results at each run.
\n\"rerun\" should be get our saved checkpoint in 2 final submission then re-predict updated private test set, not rerun the entire notebook. You should update private test set at the time of this discussion https://www.kaggle.com/c/global-wheat-detection/discussion/163291. I spent the whole month just to choose the right submission of the same notebook.
Sorry, maybe I was wrong about the \"rerun\" concept, you just run the evaluation script again on the updated ground-truth and our final submission.csv.
", "rawMarkdown": "Sorry, maybe I was wrong about the \"rerun\" concept, you just run the evaluation script again on the updated ground-truth and our final submission.csv.", "votes": 4 }, { "id": 959966, "postDate": "2020-08-06T04:08:34.347Z", "content": "Yes, I think they will simply rerun the evaluation script\nWhile this is not a big suprize with all the roller-coaster wheat ride, but I think this should have been informed to Kagglers.
\"<20 images in test that are not the same size as the others\" at this discussion https://www.kaggle.com/c/global-wheat-detection/discussion/163291
\n\nIf you correct resolution of 20 images to 1024x1024(and ground-truth), please rescale 20 predictions of us on these images. Because we predict boxes coordinates according to the resolution of the image, not 1024x1024
", "rawMarkdown": "\"<20 images in test that are not the same size as the others\" at this discussion https://www.kaggle.com/c/global-wheat-detection/discussion/163291\n\nIf you correct resolution of 20 images to 1024x1024(and ground-truth), please rescale 20 predictions of us on these images. Because we predict boxes coordinates according to the resolution of the image, not 1024x1024", "votes": 5 }, { "id": 959989, "postDate": "2020-08-06T04:21:51.180Z", "content": "@nguyenbadung \nthat is true, since most of the top teams should be rescaling images based on the former image size..
", "rawMarkdown": "@nguyenbadung \nthat is true, since most of the top teams should be rescaling images based on the former image size..", "votes": 2 }, { "id": 960059, "postDate": "2020-08-06T05:36:48.930Z", "content": "If the rerun evaluation is correct, the private score should be +- 0,00089.
", "rawMarkdown": "If the rerun evaluation is correct, the private score should be +- 0,00089. ", "votes": 1 } ] }, { "id": 959533, "postDate": "2020-08-05T17:06:16.407Z", "content": "Thanks everyone for your participation in the Global Wheat Detection Competition!
\n\nPlease note that the Private leaderboard will be rerun with the updated private test set, so please stay tuned for the finalized leaderboard within the next couple days.
\n\nI also wanted to remind everyone that the deadline for paper submissions to the Computer Vision Problems in Plant Phenotyping workshop is Aug. 9 on CMT (https://cmt3.research.microsoft.com/GWC2020). If you plan to submit a paper, please send a short email to ian.stavness@usask.ca to let us know.
\n\nFollowing discussion in the forums, we understand some participants have re-labeled the training datasets. If you are willing to share these annotations, please email ian.stavness@usask.ca This will be a big help for us to refine the annotations for future competitions.
", "rawMarkdown": "Thanks everyone for your participation in the Global Wheat Detection Competition!\n \nPlease note that the Private leaderboard will be rerun with the updated private test set, so please stay tuned for the finalized leaderboard within the next couple days.\n\nI also wanted to remind everyone that the deadline for paper submissions to the Computer Vision Problems in Plant Phenotyping workshop is Aug. 9 on CMT (https://cmt3.research.microsoft.com/GWC2020). If you plan to submit a paper, please send a short email to ian.stavness@usask.ca to let us know.\n\nFollowing discussion in the forums, we understand some participants have re-labeled the training datasets. If you are willing to share these annotations, please email ian.stavness@usask.ca This will be a big help for us to refine the annotations for future competitions.\n \n", "votes": 8 }, { "id": 960176, "postDate": "2020-08-06T07:36:42.107Z", "content": "I think we should show more patience and caution before the task they have in front of them, do not let the thoughts fly away too long before one know the answer. Kaggle are very consistent, previous competitions with compliant rules have been handled thoroughly and for example the latest one they went through over 700 submissions and cleared away over 200, so why change their own routines now? It would confuse even the most experienced.\n\"We reviewed more than 700 Kernels, in order to make sure everyone who gets a medal has a legitimate submission\"
\n\nWait and see is a good strategy, and act on what you see when you see it. I jump into a new competition in the meantime, there is a lot to think about there, no break or rest π€ π π΄ π
", "rawMarkdown": "I think we should show more patience and caution before the task they have in front of them, do not let the thoughts fly away too long before one know the answer. Kaggle are very consistent, previous competitions with compliant rules have been handled thoroughly and for example the latest one they went through over 700 submissions and cleared away over 200, so why change their own routines now? It would confuse even the most experienced.\n\"We reviewed more than 700 Kernels, in order **to make sure everyone who gets a medal has a legitimate submission**\"\n\nWait and see is a good strategy, and act on what you see when you see it. I jump into a new competition in the meantime, there is a lot to think about there, no break or rest π€ π π΄ π ", "votes": 5 }, { "id": 959642, "postDate": "2020-08-05T19:12:23.113Z", "content": "Thanks for hosting Ian and I hope it leads to some helpful new insights and great solutions. Having setup a predictive challenge myself, I know that unexpected issues arise but glad to hear you are working for fair outcomes and planning future competitions.
", "rawMarkdown": "Thanks for hosting Ian and I hope it leads to some helpful new insights and great solutions. Having setup a predictive challenge myself, I know that unexpected issues arise but glad to hear you are working for fair outcomes and planning future competitions.", "votes": 2 }, { "id": 964259, "postDate": "2020-08-09T17:38:53.280Z", "content": "Hi,
\nThe challenge has ended.
\nBut the private test dataset's ground truth isn't available anywhere.
\nI cannot submit results now and wish to submit paper in the workshop.
\nPlease help me out.
\nThanks.
Agreed and I'm sorry you weren't able to get these results in time... The organizers should stick to their word when they same something :(\nhttps://www.kaggle.com/c/global-wheat-detection/discussion/168693#943211
", "rawMarkdown": "Agreed and I'm sorry you weren't able to get these results in time... The organizers should stick to their word when they same something :(\nhttps://www.kaggle.com/c/global-wheat-detection/discussion/168693#943211 " } ] }, { "id": 960195, "postDate": "2020-08-06T07:59:40.427Z", "rawMarkdown": "", "votes": 8, "isDeleted": true }, { "id": 960008, "postDate": "2020-08-06T04:43:17.533Z", "rawMarkdown": "", "votes": 16, "isDeleted": true, "replies": [ { "id": 960042, "postDate": "2020-08-06T05:19:12.640Z", "content": "@billarshad Totally agree. If only winners' solutions are being evaluated, non compliant submissions will get medals.
", "rawMarkdown": "@billarshad Totally agree. If only winners' solutions are being evaluated, non compliant submissions will get medals.", "votes": 4 }, { "id": 960143, "postDate": "2020-08-06T07:11:58.107Z", "content": "Totally agree. Atleast kaggle should give a definite answer.
", "rawMarkdown": "Totally agree. Atleast kaggle should give a definite answer. " }, { "id": 960178, "postDate": "2020-08-06T07:38:29.947Z", "content": "@billarshad - Totally agree. There are many of them who had better YoloV5 submissions, didn't select it due to compliance and there would be few who might have(\"Disclaimer\"), for medals. That seem to be unfair to folks who had obeyed the rules and selected compliant submission. Kaggle team should evaluate all 2270 teams submissions
", "rawMarkdown": "@billarshad - Totally agree. There are many of them who had better YoloV5 submissions, didn't select it due to compliance and there would be few who might have(\"Disclaimer\"), for medals. That seem to be unfair to folks who had obeyed the rules and selected compliant submission. Kaggle team should evaluate all 2270 teams submissions", "votes": 7 } ] }, { "id": 959719, "postDate": "2020-08-05T20:56:24.273Z", "rawMarkdown": "", "votes": -7, "isDeleted": true } ], "comments": [ { "id": 962179, "author_name": "Ian Stavness", "author_url": "", "post_date": "2020-08-07T22:01:30.697000", "content": "We are working on the finalized leaderboard, but it will still take a few more days. Thanks everyone for your patience.
", "votes": 20, "replies": [ { "id": 962506, "author_name": "xiaopeng", "author_url": "", "post_date": "2020-08-08T07:17:08.870000", "content": "Will you remove all non-compliant submissions from leaderboard?
", "votes": 1, "replies": [] }, { "id": 962516, "author_name": "Vadim Timakin", "author_url": "", "post_date": "2020-08-08T07:29:05.513000", "content": "It seems like they will. However, the question is how many places (or what part of the LB) will be checked. it's possible that after removing a lot of Yolov5 users from the medal zone this will fill up with the non-compliant submissions again so they shall check people in this zone again.
", "votes": 1, "replies": [] } ] }, { "id": 959542, "author_name": "Matt", "author_url": "", "post_date": "2020-08-05T17:17:36.430000", "content": "Thanks Ian, I did not realize this was not the corrected test set.
\n\nCould you please answer if submissions will be reviewed for license compliance? This has been a big issue in this competition with YoloV5 and it would be unfair after all this if those who abided by the rules were stuck at these standings.
", "votes": 11, "replies": [ { "id": 959662, "author_name": "Ian Stavness", "author_url": "", "post_date": "2020-08-05T19:44:49.467000", "content": "Yes, Kaggle has a standard procedure for finalizing the leaderboard and verifying compliance of winning submissions, which we are following.
", "votes": -15, "replies": [] }, { "id": 959691, "author_name": "Vadim Timakin", "author_url": "", "post_date": "2020-08-05T20:26:39.420000", "content": "Hello! What do you mean by winning solutions? Will you check only gold medal submits or the entire medal zone as well?
", "votes": 14, "replies": [] }, { "id": 959775, "author_name": "Darkate", "author_url": "", "post_date": "2020-08-05T22:04:19.743000", "content": "Verifying compliance of winning submissions only????? This is ridiculous
", "votes": 19, "replies": [] }, { "id": 959810, "author_name": "Benlei Cui", "author_url": "", "post_date": "2020-08-05T23:04:25.100000", "content": "The kaggle team worker has told us that we canβt get a medal by YOLO V5 (as the discussion mentioned ) resulting in most of us didn't choosed YOLO V5 as final submission.So i believe kaggle must can give us a fair result :D
", "votes": 19, "replies": [] }, { "id": 959887, "author_name": "Wonjun Park", "author_url": "", "post_date": "2020-08-06T02:08:19.297000", "content": "If you check only a few kernels in the top tier, it is a deception for all participants who have followed the rules.
", "votes": 11, "replies": [] }, { "id": 960089, "author_name": "pranshu", "author_url": "", "post_date": "2020-08-06T06:09:53.280000", "content": "As many have asked what is really a \"winning solution\" olny gold region or the whole medal zone ??\nCan you shed some light\n@stavness \nIs this a fair with every one who obeyed the rules \nNeed some answers
", "votes": 1, "replies": [] }, { "id": 960305, "author_name": "Jaideep", "author_url": "", "post_date": "2020-08-06T09:47:08.507000", "content": "i wish i would have selected yolo v5 then would have got medal. ILooks like Medal positions may remain stand still this is just for winning solution.
", "votes": -1, "replies": [] }, { "id": 960323, "author_name": "Ahmmad Rashid", "author_url": "", "post_date": "2020-08-06T10:03:06.420000", "content": "We trusted kaggle and submitted complient solution
", "votes": 5, "replies": [] } ] }, { "id": 959923, "author_name": "DungNB", "author_url": "", "post_date": "2020-08-06T03:08:06.607000", "content": "This is ridiculous. Because we train on hidden testset with augmentation, we get different results at each run.
\n\"rerun\" should be get our saved checkpoint in 2 final submission then re-predict updated private test set, not rerun the entire notebook. You should update private test set at the time of this discussion https://www.kaggle.com/c/global-wheat-detection/discussion/163291. I spent the whole month just to choose the right submission of the same notebook.
Sorry, maybe I was wrong about the \"rerun\" concept, you just run the evaluation script again on the updated ground-truth and our final submission.csv.
", "votes": 4, "replies": [] }, { "id": 959966, "author_name": "arutema47", "author_url": "", "post_date": "2020-08-06T04:08:34.347000", "content": "Yes, I think they will simply rerun the evaluation script\nWhile this is not a big suprize with all the roller-coaster wheat ride, but I think this should have been informed to Kagglers.
\"<20 images in test that are not the same size as the others\" at this discussion https://www.kaggle.com/c/global-wheat-detection/discussion/163291
\n\nIf you correct resolution of 20 images to 1024x1024(and ground-truth), please rescale 20 predictions of us on these images. Because we predict boxes coordinates according to the resolution of the image, not 1024x1024
", "votes": 5, "replies": [] }, { "id": 959989, "author_name": "arutema47", "author_url": "", "post_date": "2020-08-06T04:21:51.180000", "content": "@nguyenbadung \nthat is true, since most of the top teams should be rescaling images based on the former image size..
", "votes": 2, "replies": [] }, { "id": 960059, "author_name": "DungNB", "author_url": "", "post_date": "2020-08-06T05:36:48.930000", "content": "If the rerun evaluation is correct, the private score should be +- 0,00089.
", "votes": 1, "replies": [] } ] }, { "id": 960176, "author_name": "Kirderf", "author_url": "", "post_date": "2020-08-06T07:36:42.107000", "content": "I think we should show more patience and caution before the task they have in front of them, do not let the thoughts fly away too long before one know the answer. Kaggle are very consistent, previous competitions with compliant rules have been handled thoroughly and for example the latest one they went through over 700 submissions and cleared away over 200, so why change their own routines now? It would confuse even the most experienced.\n\"We reviewed more than 700 Kernels, in order to make sure everyone who gets a medal has a legitimate submission\"
\n\nWait and see is a good strategy, and act on what you see when you see it. I jump into a new competition in the meantime, there is a lot to think about there, no break or rest π€ π π΄ π
", "votes": 5, "replies": [] }, { "id": 959642, "author_name": "MichaelP", "author_url": "", "post_date": "2020-08-05T19:12:23.113000", "content": "Thanks for hosting Ian and I hope it leads to some helpful new insights and great solutions. Having setup a predictive challenge myself, I know that unexpected issues arise but glad to hear you are working for fair outcomes and planning future competitions.
", "votes": 2, "replies": [] }, { "id": 964259, "author_name": "Neo Blue", "author_url": "", "post_date": "2020-08-09T17:38:53.280000", "content": "Hi,
\nThe challenge has ended.
\nBut the private test dataset's ground truth isn't available anywhere.
\nI cannot submit results now and wish to submit paper in the workshop.
\nPlease help me out.
\nThanks.
Agreed and I'm sorry you weren't able to get these results in time... The organizers should stick to their word when they same something :(\nhttps://www.kaggle.com/c/global-wheat-detection/discussion/168693#943211
", "votes": 0, "replies": [] } ] }, { "id": 960195, "author_name": "", "author_url": "", "post_date": "2020-08-06T07:59:40.427000", "content": "", "votes": 8, "replies": [] }, { "id": 960008, "author_name": "", "author_url": "", "post_date": "2020-08-06T04:43:17.533000", "content": "", "votes": 16, "replies": [ { "id": 960042, "author_name": "Kaushal Shah", "author_url": "", "post_date": "2020-08-06T05:19:12.640000", "content": "@billarshad Totally agree. If only winners' solutions are being evaluated, non compliant submissions will get medals.
", "votes": 4, "replies": [] }, { "id": 960143, "author_name": "Rifat Rashid", "author_url": "", "post_date": "2020-08-06T07:11:58.107000", "content": "Totally agree. Atleast kaggle should give a definite answer.
", "votes": 0, "replies": [] }, { "id": 960178, "author_name": "Manoj Prabhakar", "author_url": "", "post_date": "2020-08-06T07:38:29.947000", "content": "@billarshad - Totally agree. There are many of them who had better YoloV5 submissions, didn't select it due to compliance and there would be few who might have(\"Disclaimer\"), for medals. That seem to be unfair to folks who had obeyed the rules and selected compliant submission. Kaggle team should evaluate all 2270 teams submissions
", "votes": 7, "replies": [] } ] }, { "id": 959719, "author_name": "", "author_url": "", "post_date": "2020-08-05T20:56:24.273000", "content": "", "votes": -7, "replies": [] } ], "raw_markdown_by_id": { "962179": "We are working on the finalized leaderboard, but it will still take a few more days. Thanks everyone for your patience.", "959542": "Thanks Ian, I did not realize this was not the corrected test set.\n\nCould you please answer if submissions will be reviewed for license compliance? This has been a big issue in this competition with YoloV5 and it would be unfair after all this if those who abided by the rules were stuck at these standings. ", "959923": "This is ridiculous. Because we train on hidden testset with augmentation, we get different results at each run. \n\"rerun\" should be get our saved checkpoint in 2 final submission then re-predict updated private test set, not rerun the entire notebook. You should update private test set at the time of this discussion https://www.kaggle.com/c/global-wheat-detection/discussion/163291. I spent the whole month just to choose the right submission of the same notebook.", "959533": "Thanks everyone for your participation in the Global Wheat Detection Competition!\n \nPlease note that the Private leaderboard will be rerun with the updated private test set, so please stay tuned for the finalized leaderboard within the next couple days.\n\nI also wanted to remind everyone that the deadline for paper submissions to the Computer Vision Problems in Plant Phenotyping workshop is Aug. 9 on CMT (https://cmt3.research.microsoft.com/GWC2020). If you plan to submit a paper, please send a short email to ian.stavness@usask.ca to let us know.\n\nFollowing discussion in the forums, we understand some participants have re-labeled the training datasets. If you are willing to share these annotations, please email ian.stavness@usask.ca This will be a big help for us to refine the annotations for future competitions.\n \n", "960176": "I think we should show more patience and caution before the task they have in front of them, do not let the thoughts fly away too long before one know the answer. Kaggle are very consistent, previous competitions with compliant rules have been handled thoroughly and for example the latest one they went through over 700 submissions and cleared away over 200, so why change their own routines now? It would confuse even the most experienced.\n\"We reviewed more than 700 Kernels, in order **to make sure everyone who gets a medal has a legitimate submission**\"\n\nWait and see is a good strategy, and act on what you see when you see it. I jump into a new competition in the meantime, there is a lot to think about there, no break or rest π€ π π΄ π ", "959642": "Thanks for hosting Ian and I hope it leads to some helpful new insights and great solutions. Having setup a predictive challenge myself, I know that unexpected issues arise but glad to hear you are working for fair outcomes and planning future competitions.", "964259": "Hi, \nThe challenge has ended.\nBut the private test dataset's ground truth isn't available anywhere. \nI cannot submit results now and wish to submit paper in the workshop.\nPlease help me out.\nThanks.", "960195": "", "960008": "", "959719": "" } }