{
  "id": 53541,
  "title": "Has anyone successfully run tensorflow official mask rcnn ?",
  "url": "/competitions/data-science-bowl-2018/discussion/53541",
  "author_name": "",
  "post_date": "2018-04-01T09:08:39.350797500Z",
  "votes": null,
  "comment_count": 1,
  "views": 0,
  "content": "<p><a href=\"https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/instance_segmentation.md\">tensorflow official mask rcnn</a> <br>\nI tried but without success, and it used too much memory. Eager to know if some one has good result from tensorflow official mask rcnn.</p>",
  "messages": [
    {
      "id": "307274",
      "postDate": "04/01/2018 09:08:39",
      "content": "<p><a href=\"https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/instance_segmentation.md\">tensorflow official mask rcnn</a> <br>\nI tried but without success, and it used too much memory. Eager to know if some one has good result from tensorflow official mask rcnn.</p>",
      "rawMarkdown": "[tensorflow official mask rcnn][1]  \nI tried but without success, and it used too much memory. Eager to know if some one has good result from tensorflow official mask rcnn.\n\n\n  [1]: https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/instance_segmentation.md",
      "votes": null
    },
    {
      "id": "307344",
      "postDate": "04/01/2018 12:36:13",
      "content": "<p>Nope I couldn't make it run on a custom dataset. Plus for any configuration if you change the default batch_size from 1 to 2 things again break. And the fact that they wrote everything in slim makes it even harder to debug or follow their logic.</p>",
      "rawMarkdown": "Nope I couldn't make it run on a custom dataset. Plus for any configuration if you change the default batch_size from 1 to 2 things again break. And the fact that they wrote everything in slim makes it even harder to debug or follow their logic.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 307344,
      "author_name": "kirk86",
      "author_url": "",
      "post_date": "04/01/2018 12:36:13",
      "content": "<p>Nope I couldn't make it run on a custom dataset. Plus for any configuration if you change the default batch_size from 1 to 2 things again break. And the fact that they wrote everything in slim makes it even harder to debug or follow their logic.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "307274": "[tensorflow official mask rcnn][1]  \nI tried but without success, and it used too much memory. Eager to know if some one has good result from tensorflow official mask rcnn.\n\n\n  [1]: https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/instance_segmentation.md",
    "307344": "Nope I couldn't make it run on a custom dataset. Plus for any configuration if you change the default batch_size from 1 to 2 things again break. And the fact that they wrote everything in slim makes it even harder to debug or follow their logic."
  },
  "source": "meta"
}