{
  "id": 208410,
  "title": "Competition metric calculator",
  "url": "/competitions/vinbigdata-chest-xray-abnormalities-detection/discussion/208410",
  "author_name": "Peter",
  "post_date": "2021-01-03T10:45:37.987000",
  "votes": 18,
  "comment_count": 2,
  "views": 0,
  "content": "<p>I've just published a small metric-calculator class. You can find it <a href=\"https://www.kaggle.com/pestipeti/competition-metric-map-0-4\" target=\"_blank\">here</a>.</p>\n<p>So far, I've only tested with empty (class=14) predictions, so there might be minor bugs. Leave me a comment if you find any.</p>\n<p><strong>Notes:</strong></p>\n<p>Class 14 (no-finding) seems stable, but the public LB baseline (0.052) is slightly higher than the average.</p>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F864684%2Fc6fbe8ebcf6a481a9661c7a34142a5f0%2Fresults.png?generation=1609670701379657&amp;alt=media\" alt=\"\"></p>\n<p>We have only 300 images in the public LB, so there might be missing classes (I haven't checked it yet)</p>",
  "messages": [
    {
      "id": 1136696,
      "postDate": "2021-01-03T10:45:37.987Z",
      "content": "<p>I've just published a small metric-calculator class. You can find it <a href=\"https://www.kaggle.com/pestipeti/competition-metric-map-0-4\" target=\"_blank\">here</a>.</p>\n<p>So far, I've only tested with empty (class=14) predictions, so there might be minor bugs. Leave me a comment if you find any.</p>\n<p><strong>Notes:</strong></p>\n<p>Class 14 (no-finding) seems stable, but the public LB baseline (0.052) is slightly higher than the average.</p>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F864684%2Fc6fbe8ebcf6a481a9661c7a34142a5f0%2Fresults.png?generation=1609670701379657&amp;alt=media\" alt=\"\"></p>\n<p>We have only 300 images in the public LB, so there might be missing classes (I haven't checked it yet)</p>",
      "rawMarkdown": "I've just published a small metric-calculator class. You can find it [here](https://www.kaggle.com/pestipeti/competition-metric-map-0-4).\n\nSo far, I've only tested with empty (class=14) predictions, so there might be minor bugs. Leave me a comment if you find any.\n\n**Notes:**\n\nClass 14 (no-finding) seems stable, but the public LB baseline (0.052) is slightly higher than the average.\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F864684%2Fc6fbe8ebcf6a481a9661c7a34142a5f0%2Fresults.png?generation=1609670701379657&alt=media)\n\nWe have only 300 images in the public LB, so there might be missing classes (I haven't checked it yet)\n\n\n",
      "votes": 18
    },
    {
      "id": 1137910,
      "postDate": "2021-01-04T09:23:27.903Z",
      "content": "<p>Great Initiative and I hope you will update it for more classes.<br>\nThanks</p>",
      "rawMarkdown": "Great Initiative and I hope you will update it for more classes.\nThanks"
    },
    {
      "id": 1136864,
      "postDate": "2021-01-03T13:54:05.907Z",
      "content": "<p>I think this correlates with LB. can we integrate this while training so that we can do loss.forward instead of loss.backward. and optimizing to a maximum score</p>",
      "rawMarkdown": "I think this correlates with LB. can we integrate this while training so that we can do loss.forward instead of loss.backward. and optimizing to a maximum score"
    }
  ],
  "comments": [
    {
      "id": 1137910,
      "author_name": "Arpit Bhushan Sharma",
      "author_url": "",
      "post_date": "2021-01-04T09:23:27.903000",
      "content": "<p>Great Initiative and I hope you will update it for more classes.<br>\nThanks</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 1136864,
      "author_name": "Mohammed Rizin V K",
      "author_url": "",
      "post_date": "2021-01-03T13:54:05.907000",
      "content": "<p>I think this correlates with LB. can we integrate this while training so that we can do loss.forward instead of loss.backward. and optimizing to a maximum score</p>",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1136696": "I've just published a small metric-calculator class. You can find it [here](https://www.kaggle.com/pestipeti/competition-metric-map-0-4).\n\nSo far, I've only tested with empty (class=14) predictions, so there might be minor bugs. Leave me a comment if you find any.\n\n**Notes:**\n\nClass 14 (no-finding) seems stable, but the public LB baseline (0.052) is slightly higher than the average.\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F864684%2Fc6fbe8ebcf6a481a9661c7a34142a5f0%2Fresults.png?generation=1609670701379657&alt=media)\n\nWe have only 300 images in the public LB, so there might be missing classes (I haven't checked it yet)\n\n\n",
    "1137910": "Great Initiative and I hope you will update it for more classes.\nThanks",
    "1136864": "I think this correlates with LB. can we integrate this while training so that we can do loss.forward instead of loss.backward. and optimizing to a maximum score"
  }
}