{
  "id": 298001,
  "title": "Some curiosities after competition ends",
  "url": "/competitions/sartorius-cell-instance-segmentation/discussion/298001",
  "author_name": "",
  "post_date": "2021-12-31T03:26:46.862268800Z",
  "votes": null,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Although competition is already ended, I still have a quick question and want to hear what do you think about it. It would be super helpful if you could share your thought with me!</p>\n<p>SWIN-TRANSFORMER!!! Is there any top solution using this backbone? If so, how did it perform? From my experience, the performance is not as good as resnet even when adopting Swin-L as backbone. Any thoughts or explainations about why it didn't work? Also the ResNeSt is not working for me too while in the LIVECell official paper, this worked pretty well!</p>\n<p>Happy new year!!! Thanks for this period's accompany.</p>",
  "messages": [
    {
      "id": "1633761",
      "postDate": "12/31/2021 03:26:46",
      "content": "<p>Although competition is already ended, I still have a quick question and want to hear what do you think about it. It would be super helpful if you could share your thought with me!</p>\n<p>SWIN-TRANSFORMER!!! Is there any top solution using this backbone? If so, how did it perform? From my experience, the performance is not as good as resnet even when adopting Swin-L as backbone. Any thoughts or explainations about why it didn't work? Also the ResNeSt is not working for me too while in the LIVECell official paper, this worked pretty well!</p>\n<p>Happy new year!!! Thanks for this period's accompany.</p>",
      "rawMarkdown": "Although competition is already ended, I still have a quick question and want to hear what do you think about it. It would be super helpful if you could share your thought with me!\n\nSWIN-TRANSFORMER!!! Is there any top solution using this backbone? If so, how did it perform? From my experience, the performance is not as good as resnet even when adopting Swin-L as backbone. Any thoughts or explainations about why it didn't work? Also the ResNeSt is not working for me too while in the LIVECell official paper, this worked pretty well!\n\nHappy new year!!! Thanks for this period's accompany.",
      "votes": null
    },
    {
      "id": "1633770",
      "postDate": "12/31/2021 03:51:45",
      "content": "<p>ResNeSt did not work for me either..</p>",
      "rawMarkdown": "ResNeSt did not work for me either..",
      "votes": null
    },
    {
      "id": "1633779",
      "postDate": "12/31/2021 04:03:33",
      "content": "<p>Single fold model with MaskRCNN Swin Tiny Transformer (mmdetection) can get 0.334 LB (not optimized)</p>",
      "rawMarkdown": "Single fold model with MaskRCNN Swin Tiny Transformer (mmdetection) can get 0.334 LB (not optimized)",
      "votes": null
    },
    {
      "id": "1633806",
      "postDate": "12/31/2021 04:41:20",
      "content": "<p>Impressive! I guess maybe it’s due to some adoption mistakes on my side!</p>",
      "rawMarkdown": "Impressive! I guess maybe it’s due to some adoption mistakes on my side!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1633770,
      "author_name": "yousof9",
      "author_url": "",
      "post_date": "12/31/2021 03:51:45",
      "content": "<p>ResNeSt did not work for me either..</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1633779,
      "author_name": "duykhanh99",
      "author_url": "",
      "post_date": "12/31/2021 04:03:33",
      "content": "<p>Single fold model with MaskRCNN Swin Tiny Transformer (mmdetection) can get 0.334 LB (not optimized)</p>",
      "votes": null,
      "replies": [
        {
          "id": 1633806,
          "author_name": "charonwangg",
          "author_url": "",
          "post_date": "12/31/2021 04:41:20",
          "content": "<p>Impressive! I guess maybe it’s due to some adoption mistakes on my side!</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1633761": "Although competition is already ended, I still have a quick question and want to hear what do you think about it. It would be super helpful if you could share your thought with me!\n\nSWIN-TRANSFORMER!!! Is there any top solution using this backbone? If so, how did it perform? From my experience, the performance is not as good as resnet even when adopting Swin-L as backbone. Any thoughts or explainations about why it didn't work? Also the ResNeSt is not working for me too while in the LIVECell official paper, this worked pretty well!\n\nHappy new year!!! Thanks for this period's accompany.",
    "1633770": "ResNeSt did not work for me either..",
    "1633779": "Single fold model with MaskRCNN Swin Tiny Transformer (mmdetection) can get 0.334 LB (not optimized)",
    "1633806": "Impressive! I guess maybe it’s due to some adoption mistakes on my side!"
  },
  "source": "meta"
}