{
  "id": 465772,
  "title": "0.0000 submission score",
  "url": "/competitions/blood-vessel-segmentation/discussion/465772",
  "author_name": "",
  "post_date": "2024-01-05T14:47:56.809011800Z",
  "votes": null,
  "comment_count": 4,
  "views": 0,
  "content": "<p>hey guys<br>\nI'm experiencing score of 0.0000 which i think shouldn't be. Model has validation IoU coefficient of 0.9475 on about 45% of the train images and the corresponding dice coefficient is 0.9457 but getting score of 0.0000 despite increasing the threshold to about 85% confidence. I also resized the masks to the shape of the input images before rle encoding. what could be the cause.<br>\nThis is the notebook link for anyone that wishes to help <a href=\"https://www.kaggle.com/code/b45370/inferencenb\" target=\"_blank\">https://www.kaggle.com/code/b45370/inferencenb</a><br>\nwith regards. Happy coding</p>",
  "messages": [
    {
      "id": "2588510",
      "postDate": "01/05/2024 14:47:56",
      "content": "<p>hey guys<br>\nI'm experiencing score of 0.0000 which i think shouldn't be. Model has validation IoU coefficient of 0.9475 on about 45% of the train images and the corresponding dice coefficient is 0.9457 but getting score of 0.0000 despite increasing the threshold to about 85% confidence. I also resized the masks to the shape of the input images before rle encoding. what could be the cause.<br>\nThis is the notebook link for anyone that wishes to help <a href=\"https://www.kaggle.com/code/b45370/inferencenb\" target=\"_blank\">https://www.kaggle.com/code/b45370/inferencenb</a><br>\nwith regards. Happy coding</p>",
      "rawMarkdown": "hey guys\nI'm experiencing score of 0.0000 which i think shouldn't be. Model has validation IoU coefficient of 0.9475 on about 45% of the train images and the corresponding dice coefficient is 0.9457 but getting score of 0.0000 despite increasing the threshold to about 85% confidence. I also resized the masks to the shape of the input images before rle encoding. what could be the cause.\nThis is the notebook link for anyone that wishes to help https://www.kaggle.com/code/b45370/inferencenb\nwith regards. Happy coding",
      "votes": null
    },
    {
      "id": "2588727",
      "postDate": "01/05/2024 17:17:21",
      "content": "<p>Well. As an example the local test images are empty masks and your predictions are not.<br>\n0    kidney_5_0000   559634 2 560546 2 561458 2 562349 7 563261 7 5…<br>\n1    kidney_5_0001   30845 2 31757 2 32669 2 330095 2 331007 2 3319…<br>\n2    kidney_5_0002   259953 2 260865 2 261777 2 314351 1 315263 1 3…<br>\n3    kidney_6_0000   559634 2 560546 2 561458 2 562349 7 563261 7 5…<br>\n4    kidney_6_0001   30845 2 31757 2 32669 2 330095 2 331007 2 3319…<br>\nSo something is going wrong.</p>",
      "rawMarkdown": "Well. As an example the local test images are empty masks and your predictions are not.\n0\tkidney_5_0000\t559634 2 560546 2 561458 2 562349 7 563261 7 5...\n1\tkidney_5_0001\t30845 2 31757 2 32669 2 330095 2 331007 2 3319...\n2\tkidney_5_0002\t259953 2 260865 2 261777 2 314351 1 315263 1 3...\n3\tkidney_6_0000\t559634 2 560546 2 561458 2 562349 7 563261 7 5...\n4\tkidney_6_0001\t30845 2 31757 2 32669 2 330095 2 331007 2 3319...\nSo something is going wrong.",
      "votes": null
    },
    {
      "id": "2588760",
      "postDate": "01/05/2024 17:52:52",
      "content": "<p>Oh. Okay. I'll recheck to see what's going on. but visualizing the output, i noticed no pixel has value of 1. maybe it's just small</p>",
      "rawMarkdown": "Oh. Okay. I'll recheck to see what's going on. but visualizing the output, i noticed no pixel has value of 1. maybe it's just small",
      "votes": null
    },
    {
      "id": "2588880",
      "postDate": "01/05/2024 20:18:24",
      "content": "<p>I'm using the same approach. probably you have resizing problem too. I think tf.resize(…) is not a proper way to convert to the original size.</p>",
      "rawMarkdown": "I'm using the same approach. probably you have resizing problem too. I think tf.resize(...) is not a proper way to convert to the original size.",
      "votes": null
    },
    {
      "id": "2589034",
      "postDate": "01/06/2024 02:03:05",
      "content": "<p>okay. I'll try another resizing option. Thanks a lot</p>",
      "rawMarkdown": "okay. I'll try another resizing option. Thanks a lot",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2588727,
      "author_name": "sacuscreed",
      "author_url": "",
      "post_date": "01/05/2024 17:17:21",
      "content": "<p>Well. As an example the local test images are empty masks and your predictions are not.<br>\n0    kidney_5_0000   559634 2 560546 2 561458 2 562349 7 563261 7 5…<br>\n1    kidney_5_0001   30845 2 31757 2 32669 2 330095 2 331007 2 3319…<br>\n2    kidney_5_0002   259953 2 260865 2 261777 2 314351 1 315263 1 3…<br>\n3    kidney_6_0000   559634 2 560546 2 561458 2 562349 7 563261 7 5…<br>\n4    kidney_6_0001   30845 2 31757 2 32669 2 330095 2 331007 2 3319…<br>\nSo something is going wrong.</p>",
      "votes": null,
      "replies": [
        {
          "id": 2588760,
          "author_name": "b45370",
          "author_url": "",
          "post_date": "01/05/2024 17:52:52",
          "content": "<p>Oh. Okay. I'll recheck to see what's going on. but visualizing the output, i noticed no pixel has value of 1. maybe it's just small</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 2588880,
      "author_name": "nimaiji",
      "author_url": "",
      "post_date": "01/05/2024 20:18:24",
      "content": "<p>I'm using the same approach. probably you have resizing problem too. I think tf.resize(…) is not a proper way to convert to the original size.</p>",
      "votes": null,
      "replies": [
        {
          "id": 2589034,
          "author_name": "b45370",
          "author_url": "",
          "post_date": "01/06/2024 02:03:05",
          "content": "<p>okay. I'll try another resizing option. Thanks a lot</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2588510": "hey guys\nI'm experiencing score of 0.0000 which i think shouldn't be. Model has validation IoU coefficient of 0.9475 on about 45% of the train images and the corresponding dice coefficient is 0.9457 but getting score of 0.0000 despite increasing the threshold to about 85% confidence. I also resized the masks to the shape of the input images before rle encoding. what could be the cause.\nThis is the notebook link for anyone that wishes to help https://www.kaggle.com/code/b45370/inferencenb\nwith regards. Happy coding",
    "2588727": "Well. As an example the local test images are empty masks and your predictions are not.\n0\tkidney_5_0000\t559634 2 560546 2 561458 2 562349 7 563261 7 5...\n1\tkidney_5_0001\t30845 2 31757 2 32669 2 330095 2 331007 2 3319...\n2\tkidney_5_0002\t259953 2 260865 2 261777 2 314351 1 315263 1 3...\n3\tkidney_6_0000\t559634 2 560546 2 561458 2 562349 7 563261 7 5...\n4\tkidney_6_0001\t30845 2 31757 2 32669 2 330095 2 331007 2 3319...\nSo something is going wrong.",
    "2588760": "Oh. Okay. I'll recheck to see what's going on. but visualizing the output, i noticed no pixel has value of 1. maybe it's just small",
    "2588880": "I'm using the same approach. probably you have resizing problem too. I think tf.resize(...) is not a proper way to convert to the original size.",
    "2589034": "okay. I'll try another resizing option. Thanks a lot"
  },
  "source": "meta"
}