{
  "id": 68790,
  "title": "Make COCO Dateset for airbus challenge",
  "url": "/competitions/airbus-ship-detection/discussion/68790",
  "author_name": "pascal1129",
  "post_date": "2018-10-17T07:54:32.701000",
  "votes": 21,
  "comment_count": 19,
  "views": 0,
  "content": "<p>This code is for Kaggle competiton Airbus Ship Detection Challenge, which can help you transform RLE to coco dataset. It has 3 main functions:</p>\n\n<ol>\n<li>Delete useless images without ships;</li>\n<li>Transform RLE into coco;</li>\n<li>Store bad annotations into folder ../tmp/bad_anns;</li>\n</ol>\n\n<p>Here is my code :</p>\n\n<p><a href=\"https://github.com/pascal1129/airbus_rle_to_coco\">Github|airbus_rle_to_coco</a></p>\n\n<p>If you like this, please give me a star, thanks.</p>",
  "messages": [
    {
      "id": 405304,
      "postDate": "2018-10-17T07:54:32.700Z",
      "content": "<p>This code is for Kaggle competiton Airbus Ship Detection Challenge, which can help you transform RLE to coco dataset. It has 3 main functions:</p>\n\n<ol>\n<li>Delete useless images without ships;</li>\n<li>Transform RLE into coco;</li>\n<li>Store bad annotations into folder ../tmp/bad_anns;</li>\n</ol>\n\n<p>Here is my code :</p>\n\n<p><a href=\"https://github.com/pascal1129/airbus_rle_to_coco\">Github|airbus_rle_to_coco</a></p>\n\n<p>If you like this, please give me a star, thanks.</p>",
      "rawMarkdown": "This code is for Kaggle competiton Airbus Ship Detection Challenge, which can help you transform RLE to coco dataset. It has 3 main functions:\n\n1. Delete useless images without ships;\n2. Transform RLE into coco;\n3. Store bad annotations into folder ../tmp/bad_anns;\n\n\nHere is my code :\n\n[Github|airbus_rle_to_coco][1]\n\nIf you like this, please give me a star, thanks.\n\n\n  [1]: https://github.com/pascal1129/airbus_rle_to_coco",
      "votes": 21
    },
    {
      "id": 417289,
      "postDate": "2018-11-08T03:13:27.847Z",
      "content": "<p>不懂怎么入手起步。。。可以细聊吗，想有个基础东西先看着学习学习</p>",
      "rawMarkdown": "不懂怎么入手起步。。。可以细聊吗，想有个基础东西先看着学习学习",
      "votes": -2
    },
    {
      "id": 415071,
      "postDate": "2018-11-04T10:54:08.210Z",
      "content": "<p>超越姐姐能加个微信或者QQ聊聊吗，学校作业，转专业的不太懂老师不怎么讲，希望给点帮助</p>",
      "rawMarkdown": "超越姐姐能加个微信或者QQ聊聊吗，学校作业，转专业的不太懂老师不怎么讲，希望给点帮助",
      "votes": -1,
      "replies": [
        {
          "id": 415442,
          "postDate": "2018-11-05T06:02:02.157Z",
          "content": "<p>直接在这问吧，我也是转专业的，这些都不是问题</p>",
          "rawMarkdown": "直接在这问吧，我也是转专业的，这些都不是问题"
        }
      ]
    },
    {
      "id": 414015,
      "postDate": "2018-11-02T01:16:46.950Z",
      "content": "<p>请问有转化回来的么？</p>",
      "rawMarkdown": "请问有转化回来的么？",
      "replies": [
        {
          "id": 414099,
          "postDate": "2018-11-02T06:11:25.640Z",
          "content": "<p>转回来的代码需要跟着你的检测框架一块改，不太容易单独发出来</p>",
          "rawMarkdown": "转回来的代码需要跟着你的检测框架一块改，不太容易单独发出来"
        },
        {
          "id": 414201,
          "postDate": "2018-11-02T10:06:25.810Z",
          "content": "<p>我现在用maskrcnn跑的，结果也是coco的，但是分割的部分是这种格式，看得我有点懵\n\"segmentation\": {\n\"size\": [\n768,\n768\n],\n\"counts\": \"`eQ66ig02O1O000000000000000000000000001O00000000000000000000000000000000000000000000000000000000O2O0NbZj:\"\n},</p>",
          "rawMarkdown": "我现在用maskrcnn跑的，结果也是coco的，但是分割的部分是这种格式，看得我有点懵\n\"segmentation\": {\n\"size\": [\n768,\n768\n],\n\"counts\": \"`eQ66ig02O1O000000000000000000000000001O00000000000000000000000000000000000000000000000000000000O2O0NbZj:\"\n},"
        },
        {
          "id": 414204,
          "postDate": "2018-11-02T10:11:12.153Z",
          "content": "<p>这一块可以用pycocotool里的相关函数转换，detectron就是这么做的，建议多阅读detectron源码，改一下就能用</p>",
          "rawMarkdown": "这一块可以用pycocotool里的相关函数转换，detectron就是这么做的，建议多阅读detectron源码，改一下就能用",
          "votes": 1
        },
        {
          "id": 415683,
          "postDate": "2018-11-05T13:47:23.567Z",
          "content": "<p>你提供的正向转化是生成poly的坐标点，而他检测出来的是rle，这样理解正确么？</p>",
          "rawMarkdown": "你提供的正向转化是生成poly的坐标点，而他检测出来的是rle，这样理解正确么？"
        },
        {
          "id": 415701,
          "postDate": "2018-11-05T14:18:21.030Z",
          "content": "<p>我制作数据集的代码：整个函数输入是rle，输出是seg多边形。</p>\n\n<p>detectron输出的是什么我不知道，中间有一个阶段的结果是mask矩阵，我在这一段把mask转成rle输出，具体流程你顺着detectron源码的执行顺序看，就能理清楚。处理方法因人而异。</p>",
          "rawMarkdown": "我制作数据集的代码：整个函数输入是rle，输出是seg多边形。\n\ndetectron输出的是什么我不知道，中间有一个阶段的结果是mask矩阵，我在这一段把mask转成rle输出，具体流程你顺着detectron源码的执行顺序看，就能理清楚。处理方法因人而异。"
        }
      ]
    },
    {
      "id": 411859,
      "postDate": "2018-10-29T05:53:49.890Z",
      "content": "<p>Trained with detectron</p>\n\n<p><img src=\"https://s1.ax1x.com/2018/10/29/i2Vmh8.jpg\" alt=\"pic\"></p>",
      "rawMarkdown": "Trained with detectron\n\n![pic](https://s1.ax1x.com/2018/10/29/i2Vmh8.jpg)"
    },
    {
      "id": 409483,
      "postDate": "2018-10-24T11:27:02.153Z",
      "content": "<p>good job</p>",
      "rawMarkdown": "good job"
    },
    {
      "id": 408141,
      "postDate": "2018-10-22T11:15:43.040Z",
      "content": "<p>牛逼</p>",
      "rawMarkdown": "牛逼"
    },
    {
      "id": 405305,
      "postDate": "2018-10-17T07:59:12.877Z",
      "content": "<p>Here is the result file, which is generated by the code given.</p>",
      "rawMarkdown": "Here is the result file, which is generated by the code given."
    },
    {
      "id": 411895,
      "postDate": "2018-10-29T07:06:06.893Z",
      "rawMarkdown": "",
      "isDeleted": true
    },
    {
      "id": 407025,
      "postDate": "2018-10-20T08:22:18.803Z",
      "rawMarkdown": "",
      "isDeleted": true
    },
    {
      "id": 406552,
      "postDate": "2018-10-19T12:45:49.160Z",
      "rawMarkdown": "",
      "isDeleted": true,
      "replies": [
        {
          "id": 406559,
          "postDate": "2018-10-19T12:50:45.427Z",
          "rawMarkdown": "",
          "isDeleted": true
        }
      ]
    },
    {
      "id": 414256,
      "postDate": "2018-11-02T12:36:12.427Z",
      "content": "<p>厲害 感謝!! thanks!!</p>",
      "rawMarkdown": "厲害 感謝!! thanks!!"
    },
    {
      "id": 407026,
      "postDate": "2018-10-20T08:23:32.373Z",
      "content": "<p>Thank you, that's useful</p>",
      "rawMarkdown": "Thank you, that's useful"
    }
  ],
  "comments": [
    {
      "id": 417289,
      "author_name": "DUNT丁兆云DM课程2018王雅琪",
      "author_url": "",
      "post_date": "2018-11-08T03:13:27.847000",
      "content": "<p>不懂怎么入手起步。。。可以细聊吗，想有个基础东西先看着学习学习</p>",
      "votes": -2,
      "replies": []
    },
    {
      "id": 415071,
      "author_name": "DUNT丁兆云DM课程2018王雅琪",
      "author_url": "",
      "post_date": "2018-11-04T10:54:08.210000",
      "content": "<p>超越姐姐能加个微信或者QQ聊聊吗，学校作业，转专业的不太懂老师不怎么讲，希望给点帮助</p>",
      "votes": -1,
      "replies": [
        {
          "id": 415442,
          "author_name": "pascal1129",
          "author_url": "",
          "post_date": "2018-11-05T06:02:02.157000",
          "content": "<p>直接在这问吧，我也是转专业的，这些都不是问题</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 414015,
      "author_name": "aimhabo",
      "author_url": "",
      "post_date": "2018-11-02T01:16:46.950000",
      "content": "<p>请问有转化回来的么？</p>",
      "votes": 0,
      "replies": [
        {
          "id": 414099,
          "author_name": "pascal1129",
          "author_url": "",
          "post_date": "2018-11-02T06:11:25.640000",
          "content": "<p>转回来的代码需要跟着你的检测框架一块改，不太容易单独发出来</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 414201,
          "author_name": "aimhabo",
          "author_url": "",
          "post_date": "2018-11-02T10:06:25.810000",
          "content": "<p>我现在用maskrcnn跑的，结果也是coco的，但是分割的部分是这种格式，看得我有点懵\n\"segmentation\": {\n\"size\": [\n768,\n768\n],\n\"counts\": \"`eQ66ig02O1O000000000000000000000000001O00000000000000000000000000000000000000000000000000000000O2O0NbZj:\"\n},</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 414204,
          "author_name": "pascal1129",
          "author_url": "",
          "post_date": "2018-11-02T10:11:12.153000",
          "content": "<p>这一块可以用pycocotool里的相关函数转换，detectron就是这么做的，建议多阅读detectron源码，改一下就能用</p>",
          "votes": 1,
          "replies": []
        },
        {
          "id": 415683,
          "author_name": "aimhabo",
          "author_url": "",
          "post_date": "2018-11-05T13:47:23.567000",
          "content": "<p>你提供的正向转化是生成poly的坐标点，而他检测出来的是rle，这样理解正确么？</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 415701,
          "author_name": "pascal1129",
          "author_url": "",
          "post_date": "2018-11-05T14:18:21.030000",
          "content": "<p>我制作数据集的代码：整个函数输入是rle，输出是seg多边形。</p>\n\n<p>detectron输出的是什么我不知道，中间有一个阶段的结果是mask矩阵，我在这一段把mask转成rle输出，具体流程你顺着detectron源码的执行顺序看，就能理清楚。处理方法因人而异。</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 411859,
      "author_name": "pascal1129",
      "author_url": "",
      "post_date": "2018-10-29T05:53:49.890000",
      "content": "<p>Trained with detectron</p>\n\n<p><img src=\"https://s1.ax1x.com/2018/10/29/i2Vmh8.jpg\" alt=\"pic\"></p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 409483,
      "author_name": "ChenChch",
      "author_url": "",
      "post_date": "2018-10-24T11:27:02.153000",
      "content": "<p>good job</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 408141,
      "author_name": "Marc chen",
      "author_url": "",
      "post_date": "2018-10-22T11:15:43.040000",
      "content": "<p>牛逼</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 405305,
      "author_name": "pascal1129",
      "author_url": "",
      "post_date": "2018-10-17T07:59:12.877000",
      "content": "<p>Here is the result file, which is generated by the code given.</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 411895,
      "author_name": "",
      "author_url": "",
      "post_date": "2018-10-29T07:06:06.893000",
      "content": "",
      "votes": 0,
      "replies": []
    },
    {
      "id": 407025,
      "author_name": "",
      "author_url": "",
      "post_date": "2018-10-20T08:22:18.803000",
      "content": "",
      "votes": 0,
      "replies": []
    },
    {
      "id": 406552,
      "author_name": "",
      "author_url": "",
      "post_date": "2018-10-19T12:45:49.160000",
      "content": "",
      "votes": 0,
      "replies": [
        {
          "id": 406559,
          "author_name": "",
          "author_url": "",
          "post_date": "2018-10-19T12:50:45.427000",
          "content": "",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 414256,
      "author_name": "0xFunky",
      "author_url": "",
      "post_date": "2018-11-02T12:36:12.427000",
      "content": "<p>厲害 感謝!! thanks!!</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 407026,
      "author_name": "corner200",
      "author_url": "",
      "post_date": "2018-10-20T08:23:32.373000",
      "content": "<p>Thank you, that's useful</p>",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "405304": "This code is for Kaggle competiton Airbus Ship Detection Challenge, which can help you transform RLE to coco dataset. It has 3 main functions:\n\n1. Delete useless images without ships;\n2. Transform RLE into coco;\n3. Store bad annotations into folder ../tmp/bad_anns;\n\n\nHere is my code :\n\n[Github|airbus_rle_to_coco][1]\n\nIf you like this, please give me a star, thanks.\n\n\n  [1]: https://github.com/pascal1129/airbus_rle_to_coco",
    "417289": "不懂怎么入手起步。。。可以细聊吗，想有个基础东西先看着学习学习",
    "415071": "超越姐姐能加个微信或者QQ聊聊吗，学校作业，转专业的不太懂老师不怎么讲，希望给点帮助",
    "414015": "请问有转化回来的么？",
    "411859": "Trained with detectron\n\n![pic](https://s1.ax1x.com/2018/10/29/i2Vmh8.jpg)",
    "409483": "good job",
    "408141": "牛逼",
    "405305": "Here is the result file, which is generated by the code given.",
    "411895": "",
    "407025": "",
    "406552": "",
    "414256": "厲害 感謝!! thanks!!",
    "407026": "Thank you, that's useful"
  }
}