{
  "id": 164439,
  "title": "How to make my own train.csv file from my image dataset?",
  "url": "/competitions/global-wheat-detection/discussion/164439",
  "author_name": "",
  "post_date": "2020-07-06T08:59:26.464474800Z",
  "votes": null,
  "comment_count": 8,
  "views": 0,
  "content": "<p>Pardon me if my query is irrelevant. For my own dataset, I want to create an exact CSV file containing bounding_box, height, and weight. How should I approach it?\nIs there any kind of blogs or repositories that might help me out?</p>",
  "messages": [
    {
      "id": "917125",
      "postDate": "07/06/2020 08:59:26",
      "content": "<p>Pardon me if my query is irrelevant. For my own dataset, I want to create an exact CSV file containing bounding_box, height, and weight. How should I approach it?\nIs there any kind of blogs or repositories that might help me out?</p>",
      "rawMarkdown": "Pardon me if my query is irrelevant. For my own dataset, I want to create an exact CSV file containing bounding_box, height, and weight. How should I approach it?\nIs there any kind of blogs or repositories that might help me out?",
      "votes": null
    },
    {
      "id": "917649",
      "postDate": "07/06/2020 17:04:04",
      "content": "<p>I'm not sure I understand what you mean by your own dataset. Are you using an external dataset or something?</p>",
      "rawMarkdown": "I'm not sure I understand what you mean by your own dataset. Are you using an external dataset or something?",
      "votes": null
    },
    {
      "id": "917680",
      "postDate": "07/06/2020 17:27:33",
      "content": "<p>Yes, let's say I have accumulated a bunch of images. Now I want to extract bounding box out of them. What should be the strategy?</p>",
      "rawMarkdown": "Yes, let's say I have accumulated a bunch of images. Now I want to extract bounding box out of them. What should be the strategy?",
      "votes": null
    },
    {
      "id": "917703",
      "postDate": "07/06/2020 17:56:06",
      "content": "<p>Well first I'm curious about the source of your images, is it an external dataset or are these pictures you took yourself? I'm still not exactly sure what you mean by \"extract bounding box out of them\". Are they actually annotated with boxes or are they just images? If they're just images, you either have to annotate them yourself with something like <a href=\"https://github.com/tzutalin/labelImg\">this</a> or run your model on them to generate pseudolabels.</p>",
      "rawMarkdown": "Well first I'm curious about the source of your images, is it an external dataset or are these pictures you took yourself? I'm still not exactly sure what you mean by \"extract bounding box out of them\". Are they actually annotated with boxes or are they just images? If they're just images, you either have to annotate them yourself with something like [this](https://github.com/tzutalin/labelImg) or run your model on them to generate pseudolabels.",
      "votes": null
    },
    {
      "id": "918310",
      "postDate": "07/07/2020 07:00:02",
      "content": "<p>Oh! this is the answer I wanted to know</p>\n\n<blockquote>\n  <p>If they're just images, you either have to annotate them yourself with something like this</p>\n</blockquote>\n\n<p>Thanks for enlightening me 😊 </p>",
      "rawMarkdown": "Oh! this is the answer I wanted to know\n&gt; If they're just images, you either have to annotate them yourself with something like this\n\nThanks for enlightening me 😊",
      "votes": null
    },
    {
      "id": "918335",
      "postDate": "07/07/2020 07:28:16",
      "content": "<p>No problem. That is just one example of an image annotation tool. In this competition, the data organizers actually used a semi-supervised learning method which you may find helpful! The basic idea is covered in this graphic from <a href=\"https://arxiv.org/pdf/2005.02162.pdf\">their paper</a>:\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F846065%2F948bf0bf003ec1d78d578a04ab2d0e01%2FScreenshot%20from%202020-07-07%2003-24-55.png?generation=1594106767343398&amp;alt=media\" alt=\"\">\nAnd the tool they used to make the annotations can be found <a href=\"https://github.com/jsbroks/coco-annotator\">here</a>.</p>",
      "rawMarkdown": "No problem. That is just one example of an image annotation tool. In this competition, the data organizers actually used a semi-supervised learning method which you may find helpful! The basic idea is covered in this graphic from [their paper](https://arxiv.org/pdf/2005.02162.pdf):\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F846065%2F948bf0bf003ec1d78d578a04ab2d0e01%2FScreenshot%20from%202020-07-07%2003-24-55.png?generation=1594106767343398&amp;alt=media)\nAnd the tool they used to make the annotations can be found [here](https://github.com/jsbroks/coco-annotator).",
      "votes": null
    },
    {
      "id": "918415",
      "postDate": "07/07/2020 08:37:40",
      "content": "<p><a href=\"/naimur978\">@naimur978</a> You might want to try this too: <a href=\"https://github.com/tzutalin/labelImg\">labelImg</a></p>\n\n<p>I found it very easy to install and use. It can output both YOLO and PASCAL VOC formats (but not at the same time). You just switch to the other format and save.</p>",
      "rawMarkdown": "naimur978 You might want to try this too: [labelImg](https://github.com/tzutalin/labelImg)\n\nI found it very easy to install and use. It can output both YOLO and PASCAL VOC formats (but not at the same time). You just switch to the other format and save.",
      "votes": null
    },
    {
      "id": "919830",
      "postDate": "07/08/2020 06:06:59",
      "content": "<p><a href=\"/matthewmasters\">@matthewmasters</a> Thanks a lot for taking the time to explain this. 😍 </p>",
      "rawMarkdown": "matthewmasters Thanks a lot for taking the time to explain this. 😍",
      "votes": null
    },
    {
      "id": "919832",
      "postDate": "07/08/2020 06:09:00",
      "content": "<p><a href=\"/mperutti\">@mperutti</a> Thanks!</p>",
      "rawMarkdown": "mperutti Thanks!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 917649,
      "author_name": "matthewmasters",
      "author_url": "",
      "post_date": "07/06/2020 17:04:04",
      "content": "<p>I'm not sure I understand what you mean by your own dataset. Are you using an external dataset or something?</p>",
      "votes": null,
      "replies": [
        {
          "id": 917680,
          "author_name": "naimur978",
          "author_url": "",
          "post_date": "07/06/2020 17:27:33",
          "content": "<p>Yes, let's say I have accumulated a bunch of images. Now I want to extract bounding box out of them. What should be the strategy?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 917703,
          "author_name": "matthewmasters",
          "author_url": "",
          "post_date": "07/06/2020 17:56:06",
          "content": "<p>Well first I'm curious about the source of your images, is it an external dataset or are these pictures you took yourself? I'm still not exactly sure what you mean by \"extract bounding box out of them\". Are they actually annotated with boxes or are they just images? If they're just images, you either have to annotate them yourself with something like <a href=\"https://github.com/tzutalin/labelImg\">this</a> or run your model on them to generate pseudolabels.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 918310,
          "author_name": "naimur978",
          "author_url": "",
          "post_date": "07/07/2020 07:00:02",
          "content": "<p>Oh! this is the answer I wanted to know</p>\n\n<blockquote>\n  <p>If they're just images, you either have to annotate them yourself with something like this</p>\n</blockquote>\n\n<p>Thanks for enlightening me 😊 </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 918335,
          "author_name": "matthewmasters",
          "author_url": "",
          "post_date": "07/07/2020 07:28:16",
          "content": "<p>No problem. That is just one example of an image annotation tool. In this competition, the data organizers actually used a semi-supervised learning method which you may find helpful! The basic idea is covered in this graphic from <a href=\"https://arxiv.org/pdf/2005.02162.pdf\">their paper</a>:\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F846065%2F948bf0bf003ec1d78d578a04ab2d0e01%2FScreenshot%20from%202020-07-07%2003-24-55.png?generation=1594106767343398&amp;alt=media\" alt=\"\">\nAnd the tool they used to make the annotations can be found <a href=\"https://github.com/jsbroks/coco-annotator\">here</a>.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 918415,
          "author_name": "mperutti",
          "author_url": "",
          "post_date": "07/07/2020 08:37:40",
          "content": "<p><a href=\"/naimur978\">@naimur978</a> You might want to try this too: <a href=\"https://github.com/tzutalin/labelImg\">labelImg</a></p>\n\n<p>I found it very easy to install and use. It can output both YOLO and PASCAL VOC formats (but not at the same time). You just switch to the other format and save.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 919830,
          "author_name": "naimur978",
          "author_url": "",
          "post_date": "07/08/2020 06:06:59",
          "content": "<p><a href=\"/matthewmasters\">@matthewmasters</a> Thanks a lot for taking the time to explain this. 😍 </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 919832,
          "author_name": "naimur978",
          "author_url": "",
          "post_date": "07/08/2020 06:09:00",
          "content": "<p><a href=\"/mperutti\">@mperutti</a> Thanks!</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "917125": "Pardon me if my query is irrelevant. For my own dataset, I want to create an exact CSV file containing bounding_box, height, and weight. How should I approach it?\nIs there any kind of blogs or repositories that might help me out?",
    "917649": "I'm not sure I understand what you mean by your own dataset. Are you using an external dataset or something?",
    "917680": "Yes, let's say I have accumulated a bunch of images. Now I want to extract bounding box out of them. What should be the strategy?",
    "917703": "Well first I'm curious about the source of your images, is it an external dataset or are these pictures you took yourself? I'm still not exactly sure what you mean by \"extract bounding box out of them\". Are they actually annotated with boxes or are they just images? If they're just images, you either have to annotate them yourself with something like [this](https://github.com/tzutalin/labelImg) or run your model on them to generate pseudolabels.",
    "918310": "Oh! this is the answer I wanted to know\n&gt; If they're just images, you either have to annotate them yourself with something like this\n\nThanks for enlightening me 😊",
    "918335": "No problem. That is just one example of an image annotation tool. In this competition, the data organizers actually used a semi-supervised learning method which you may find helpful! The basic idea is covered in this graphic from [their paper](https://arxiv.org/pdf/2005.02162.pdf):\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F846065%2F948bf0bf003ec1d78d578a04ab2d0e01%2FScreenshot%20from%202020-07-07%2003-24-55.png?generation=1594106767343398&amp;alt=media)\nAnd the tool they used to make the annotations can be found [here](https://github.com/jsbroks/coco-annotator).",
    "918415": "naimur978 You might want to try this too: [labelImg](https://github.com/tzutalin/labelImg)\n\nI found it very easy to install and use. It can output both YOLO and PASCAL VOC formats (but not at the same time). You just switch to the other format and save.",
    "919830": "matthewmasters Thanks a lot for taking the time to explain this. 😍",
    "919832": "mperutti Thanks!"
  },
  "source": "meta"
}