{
  "id": 232214,
  "title": "Keras Unet Collection",
  "url": "/competitions/hubmap-kidney-segmentation/discussion/232214",
  "author_name": "",
  "post_date": "2021-04-12T17:04:38.276094Z",
  "votes": 5,
  "comment_count": 1,
  "views": 0,
  "content": "<p>A lot of Kagglers use the very fine package segmentation-models from Qubvel which provides an excellent implementation of a lot of segmentation models.</p>\n<p>Another library I came accross is <a href=\"https://github.com/yingkaisha/keras-unet-collection\" target=\"_blank\">keras-unet-collection</a>.<br>\nIt provides a wide range of Unet variants. I've tried a few networks and it works very good.</p>\n<p>So far not been able to beat the best score of what I have with segmentation-models…but that could be related to the amount of time invested yet.</p>\n<p>Anyway thought I should share this link as it is very interresting.</p>",
  "messages": [
    {
      "id": "1271502",
      "postDate": "04/12/2021 17:04:38",
      "content": "<p>A lot of Kagglers use the very fine package segmentation-models from Qubvel which provides an excellent implementation of a lot of segmentation models.</p>\n<p>Another library I came accross is <a href=\"https://github.com/yingkaisha/keras-unet-collection\" target=\"_blank\">keras-unet-collection</a>.<br>\nIt provides a wide range of Unet variants. I've tried a few networks and it works very good.</p>\n<p>So far not been able to beat the best score of what I have with segmentation-models…but that could be related to the amount of time invested yet.</p>\n<p>Anyway thought I should share this link as it is very interresting.</p>",
      "rawMarkdown": "A lot of Kagglers use the very fine package segmentation-models from Qubvel which provides an excellent implementation of a lot of segmentation models.\n\nAnother library I came accross is [keras-unet-collection](https://github.com/yingkaisha/keras-unet-collection).\nIt provides a wide range of Unet variants. I've tried a few networks and it works very good.\n\nSo far not been able to beat the best score of what I have with segmentation-models...but that could be related to the amount of time invested yet.\n\nAnyway thought I should share this link as it is very interresting.",
      "votes": null
    },
    {
      "id": "1271806",
      "postDate": "04/12/2021 23:36:34",
      "content": "<p>Thanks Robin!</p>",
      "rawMarkdown": "Thanks Robin!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1271806,
      "author_name": "cdeotte",
      "author_url": "",
      "post_date": "04/12/2021 23:36:34",
      "content": "<p>Thanks Robin!</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1271502": "A lot of Kagglers use the very fine package segmentation-models from Qubvel which provides an excellent implementation of a lot of segmentation models.\n\nAnother library I came accross is [keras-unet-collection](https://github.com/yingkaisha/keras-unet-collection).\nIt provides a wide range of Unet variants. I've tried a few networks and it works very good.\n\nSo far not been able to beat the best score of what I have with segmentation-models...but that could be related to the amount of time invested yet.\n\nAnyway thought I should share this link as it is very interresting.",
    "1271806": "Thanks Robin!"
  },
  "source": "meta"
}