{
  "id": 418084,
  "title": "How to get confidence score.",
  "url": "/competitions/hubmap-hacking-the-human-vasculature/discussion/418084",
  "author_name": "",
  "post_date": "2023-06-18T17:12:22.217213100Z",
  "votes": 1,
  "comment_count": 4,
  "views": 0,
  "content": "<p>How to get confidence score. I am using u-net model architecture using tensorflow. <br>\nI am able to predict prediction masks with good accuracy but not able to get confidence score.</p>",
  "messages": [
    {
      "id": "2308136",
      "postDate": "06/18/2023 17:12:22",
      "content": "<p>How to get confidence score. I am using u-net model architecture using tensorflow. <br>\nI am able to predict prediction masks with good accuracy but not able to get confidence score.</p>",
      "rawMarkdown": "How to get confidence score. I am using u-net model architecture using tensorflow. \nI am able to predict prediction masks with good accuracy but not able to get confidence score.",
      "votes": null
    },
    {
      "id": "2308740",
      "postDate": "06/19/2023 07:11:13",
      "content": "<p>if you use unet, you can use conf from your mask, the first you need use some thing like connected component to detection all blood in your predict masks, after this you can use average conf score of all pixel in a detection blood as your conf score.</p>",
      "rawMarkdown": "if you use unet, you can use conf from your mask, the first you need use some thing like connected component to detection all blood in your predict masks, after this you can use average conf score of all pixel in a detection blood as your conf score.",
      "votes": null
    },
    {
      "id": "2310872",
      "postDate": "06/20/2023 16:59:37",
      "content": "<p>I dint get what you said could you tell it a bit more detailed?</p>",
      "rawMarkdown": "I dint get what you said could you tell it a bit more detailed?",
      "votes": null
    },
    {
      "id": "2310959",
      "postDate": "06/20/2023 18:40:25",
      "content": "<p>Use average non-binarized mask as a confidence score.</p>",
      "rawMarkdown": "Use average non-binarized mask as a confidence score.",
      "votes": null
    },
    {
      "id": "2314153",
      "postDate": "06/23/2023 08:10:34",
      "content": "<p>As I understood he is saying take the predicted probs within the segmented part and average them up</p>",
      "rawMarkdown": "As I understood he is saying take the predicted probs within the segmented part and average them up",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2308740,
      "author_name": "hidngnguyna",
      "author_url": "",
      "post_date": "06/19/2023 07:11:13",
      "content": "<p>if you use unet, you can use conf from your mask, the first you need use some thing like connected component to detection all blood in your predict masks, after this you can use average conf score of all pixel in a detection blood as your conf score.</p>",
      "votes": null,
      "replies": [
        {
          "id": 2310872,
          "author_name": "vishakkbhat",
          "author_url": "",
          "post_date": "06/20/2023 16:59:37",
          "content": "<p>I dint get what you said could you tell it a bit more detailed?</p>",
          "votes": null,
          "replies": [
            {
              "id": 2310959,
              "author_name": "sakvaua",
              "author_url": "",
              "post_date": "06/20/2023 18:40:25",
              "content": "<p>Use average non-binarized mask as a confidence score.</p>",
              "votes": null,
              "replies": []
            },
            {
              "id": 2314153,
              "author_name": "josipvrdoljak",
              "author_url": "",
              "post_date": "06/23/2023 08:10:34",
              "content": "<p>As I understood he is saying take the predicted probs within the segmented part and average them up</p>",
              "votes": null,
              "replies": []
            }
          ]
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2308136": "How to get confidence score. I am using u-net model architecture using tensorflow. \nI am able to predict prediction masks with good accuracy but not able to get confidence score.",
    "2308740": "if you use unet, you can use conf from your mask, the first you need use some thing like connected component to detection all blood in your predict masks, after this you can use average conf score of all pixel in a detection blood as your conf score.",
    "2310872": "I dint get what you said could you tell it a bit more detailed?",
    "2310959": "Use average non-binarized mask as a confidence score.",
    "2314153": "As I understood he is saying take the predicted probs within the segmented part and average them up"
  },
  "source": "meta"
}