{
  "id": 111493,
  "title": "threshold is high?",
  "url": "/competitions/understanding_cloud_organization/discussion/111493",
  "author_name": "",
  "post_date": "2019-10-06T08:40:19.561331800Z",
  "votes": 2,
  "comment_count": 5,
  "views": 0,
  "content": "<p>When I try to get the best threshold for the post process, the best threshold is 0.95 . Wow, it is so high that I am wondering whether there is something wrong.  Looking forward for help, thanks!</p>",
  "messages": [
    {
      "id": "642532",
      "postDate": "10/06/2019 08:40:19",
      "content": "<p>When I try to get the best threshold for the post process, the best threshold is 0.95 . Wow, it is so high that I am wondering whether there is something wrong.  Looking forward for help, thanks!</p>",
      "rawMarkdown": "When I try to get the best threshold for the post process, the best threshold is 0.95 . Wow, it is so high that I am wondering whether there is something wrong.  Looking forward for help, thanks!",
      "votes": null
    },
    {
      "id": "642533",
      "postDate": "10/06/2019 08:41:18",
      "content": "<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1788154%2F13cc028d1d4b36366ce8f2f5b822695b%2F1.jpg?generation=1570351275957015&amp;alt=media\" alt=\"\"></p>",
      "rawMarkdown": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1788154%2F13cc028d1d4b36366ce8f2f5b822695b%2F1.jpg?generation=1570351275957015&amp;alt=media)",
      "votes": null
    },
    {
      "id": "642629",
      "postDate": "10/06/2019 12:13:35",
      "content": "<p>At least on my models I never got thresholds that big, also your \"min mas size\" values are a little high, but looking at your Dice scores maybe the problem is your model. </p>",
      "rawMarkdown": "At least on my models I never got thresholds that big, also your \"min mas size\" values are a little high, but looking at your Dice scores maybe the problem is your model.",
      "votes": null
    },
    {
      "id": "642659",
      "postDate": "10/06/2019 12:50:02",
      "content": "<p>Use that threshold to plot some predictions of validation set and see how it is. If the results look good to you then continue with that threshold</p>",
      "rawMarkdown": "Use that threshold to plot some predictions of validation set and see how it is. If the results look good to you then continue with that threshold",
      "votes": null
    },
    {
      "id": "653832",
      "postDate": "10/21/2019 02:52:28",
      "content": "<p>I also met the same problem! Did you solve this?</p>",
      "rawMarkdown": "I also met the same problem! Did you solve this?",
      "votes": null
    },
    {
      "id": "654094",
      "postDate": "10/21/2019 12:38:31",
      "content": "<p>Submitting all empty masks gives LB 0.477. Looks like your model does a similar thing at this threshold.\nWhich means that either you model hasn't learned anything useful or there is a bug in your validation/gridsearch pipeline.</p>",
      "rawMarkdown": "Submitting all empty masks gives LB 0.477. Looks like your model does a similar thing at this threshold.\nWhich means that either you model hasn't learned anything useful or there is a bug in your validation/gridsearch pipeline.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 642533,
      "author_name": "bigyellower",
      "author_url": "",
      "post_date": "10/06/2019 08:41:18",
      "content": "<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1788154%2F13cc028d1d4b36366ce8f2f5b822695b%2F1.jpg?generation=1570351275957015&amp;alt=media\" alt=\"\"></p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 642629,
      "author_name": "dimitreoliveira",
      "author_url": "",
      "post_date": "10/06/2019 12:13:35",
      "content": "<p>At least on my models I never got thresholds that big, also your \"min mas size\" values are a little high, but looking at your Dice scores maybe the problem is your model. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 642659,
      "author_name": "bibek777",
      "author_url": "",
      "post_date": "10/06/2019 12:50:02",
      "content": "<p>Use that threshold to plot some predictions of validation set and see how it is. If the results look good to you then continue with that threshold</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 653832,
      "author_name": "binjiezhang",
      "author_url": "",
      "post_date": "10/21/2019 02:52:28",
      "content": "<p>I also met the same problem! Did you solve this?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 654094,
      "author_name": "vzaguskin",
      "author_url": "",
      "post_date": "10/21/2019 12:38:31",
      "content": "<p>Submitting all empty masks gives LB 0.477. Looks like your model does a similar thing at this threshold.\nWhich means that either you model hasn't learned anything useful or there is a bug in your validation/gridsearch pipeline.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "642532": "When I try to get the best threshold for the post process, the best threshold is 0.95 . Wow, it is so high that I am wondering whether there is something wrong.  Looking forward for help, thanks!",
    "642533": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1788154%2F13cc028d1d4b36366ce8f2f5b822695b%2F1.jpg?generation=1570351275957015&amp;alt=media)",
    "642629": "At least on my models I never got thresholds that big, also your \"min mas size\" values are a little high, but looking at your Dice scores maybe the problem is your model.",
    "642659": "Use that threshold to plot some predictions of validation set and see how it is. If the results look good to you then continue with that threshold",
    "653832": "I also met the same problem! Did you solve this?",
    "654094": "Submitting all empty masks gives LB 0.477. Looks like your model does a similar thing at this threshold.\nWhich means that either you model hasn't learned anything useful or there is a bug in your validation/gridsearch pipeline."
  },
  "source": "meta"
}