{
  "id": 124607,
  "title": "Submission 17.26938",
  "url": "/competitions/deepfake-detection-challenge/discussion/124607",
  "author_name": "",
  "post_date": "2020-01-05T09:21:56.981697600Z",
  "votes": null,
  "comment_count": 6,
  "views": 0,
  "content": "<p>Why all my subs are getting 17.26938 score? After submitting, the score gets calculated in a minute, does not take the time it takes to run the inference kernel, If someone knows the problem that would make my day, I am wasting my subs cuz of this thing.</p>\n\n<p>I request Kaggle Team to answer, there are a lot of candidates having this score, Thank You!</p>\n\n<p>Edit: About my sub:- It is around 200 0.0 and 200 1.0 on the public validation set, even if they are not right, that should still be around random, 0.69...</p>",
  "messages": [
    {
      "id": "710831",
      "postDate": "01/05/2020 09:21:56",
      "content": "<p>Why all my subs are getting 17.26938 score? After submitting, the score gets calculated in a minute, does not take the time it takes to run the inference kernel, If someone knows the problem that would make my day, I am wasting my subs cuz of this thing.</p>\n\n<p>I request Kaggle Team to answer, there are a lot of candidates having this score, Thank You!</p>\n\n<p>Edit: About my sub:- It is around 200 0.0 and 200 1.0 on the public validation set, even if they are not right, that should still be around random, 0.69...</p>",
      "rawMarkdown": "Why all my subs are getting 17.26938 score? After submitting, the score gets calculated in a minute, does not take the time it takes to run the inference kernel, If someone knows the problem that would make my day, I am wasting my subs cuz of this thing.\n\nI request Kaggle Team to answer, there are a lot of candidates having this score, Thank You!\n\nEdit: About my sub:- It is around 200 0.0 and 200 1.0 on the public validation set, even if they are not right, that should still be around random, 0.69...",
      "votes": null
    },
    {
      "id": "710858",
      "postDate": "01/05/2020 10:43:43",
      "content": "<p>Don't use 0.0 and 1.0.\nSubmissions are scored on log loss.</p>\n\n<p><a href=\"https://www.kaggle.com/c/deepfake-detection-challenge/overview/evaluation\">https://www.kaggle.com/c/deepfake-detection-challenge/overview/evaluation</a></p>",
      "rawMarkdown": "Don't use 0.0 and 1.0.\nSubmissions are scored on log loss.\n\nhttps://www.kaggle.com/c/deepfake-detection-challenge/overview/evaluation",
      "votes": null
    },
    {
      "id": "710873",
      "postDate": "01/05/2020 11:17:50",
      "content": "<p>I even tried with regression like 0.7xxxx, 0.000xxx..., no luck</p>",
      "rawMarkdown": "I even tried with regression like 0.7xxxx, 0.000xxx..., no luck",
      "votes": null
    },
    {
      "id": "710880",
      "postDate": "01/05/2020 11:23:41",
      "content": "<h1>Case all 0</h1>\n\n<p>log_loss([\"REAL\", \"FAKE\"], [0, 0])\n-&gt; 17.269388197455342</p>\n\n<h1>Case all 1</h1>\n\n<p>log_loss([\"REAL\", \"FAKE\"], [1, 1])\n-&gt; 17.269388197455342</p>\n\n<h1>Case all right</h1>\n\n<p>log_loss([\"REAL\", \"FAKE\"], [1, 0])\n-&gt; 9.992007221626415e-16</p>\n\n<h1>Case all wrong</h1>\n\n<p>log_loss([\"REAL\", \"FAKE\"], [0, 1])\n-&gt; 34.53917619362578</p>\n\n<p>Thank You for your advise <a href=\"/lambrevam\">@lambrevam</a> but does not solve my problem...</p>",
      "rawMarkdown": "#Case all 0\nlog_loss([\"REAL\", \"FAKE\"], [0, 0])\n-&gt; 17.269388197455342\n#Case all 1\nlog_loss([\"REAL\", \"FAKE\"], [1, 1])\n-&gt; 17.269388197455342\n#Case all right\nlog_loss([\"REAL\", \"FAKE\"], [1, 0])\n-&gt; 9.992007221626415e-16\n#Case all wrong\nlog_loss([\"REAL\", \"FAKE\"], [0, 1])\n-&gt; 34.53917619362578\n\nThank You for your advise @lambrevam but does not solve my problem...",
      "votes": null
    },
    {
      "id": "711155",
      "postDate": "01/05/2020 18:47:59",
      "content": "<p>Solved by infering not over the sample_sumbission videos but over the /testvideos/* files</p>\n\n<p>credit:- <a href=\"https://www.kaggle.com/c/deepfake-detection-challenge/discussion/123260\">https://www.kaggle.com/c/deepfake-detection-challenge/discussion/123260</a></p>",
      "rawMarkdown": "Solved by infering not over the sample_sumbission videos but over the /testvideos/* files\n\ncredit:- https://www.kaggle.com/c/deepfake-detection-challenge/discussion/123260",
      "votes": null
    },
    {
      "id": "711330",
      "postDate": "01/06/2020 00:10:07",
      "content": "<p>\"Solved by infering not over the sample_sumbission videos but over the /testvideos/* files\"   what do you mean?  I have the same issue,  I only use sample_submission file for generate finalsub,  thank in advance</p>",
      "rawMarkdown": "\"Solved by infering not over the sample_sumbission videos but over the /testvideos/* files\"   what do you mean?  I have the same issue,  I only use sample_submission file for generate finalsub,  thank in advance",
      "votes": null
    },
    {
      "id": "711393",
      "postDate": "01/06/2020 03:02:43",
      "content": "<p>Hey <a href=\"/marcuslin\">@marcuslin</a>, do not use the sample sub either to iterate over videos or placing the output directly in sample submission...</p>",
      "rawMarkdown": "Hey @marcuslin, do not use the sample sub either to iterate over videos or placing the output directly in sample submission...",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 710858,
      "author_name": "lambrevam",
      "author_url": "",
      "post_date": "01/05/2020 10:43:43",
      "content": "<p>Don't use 0.0 and 1.0.\nSubmissions are scored on log loss.</p>\n\n<p><a href=\"https://www.kaggle.com/c/deepfake-detection-challenge/overview/evaluation\">https://www.kaggle.com/c/deepfake-detection-challenge/overview/evaluation</a></p>",
      "votes": null,
      "replies": [
        {
          "id": 710873,
          "author_name": "harshitsheoran",
          "author_url": "",
          "post_date": "01/05/2020 11:17:50",
          "content": "<p>I even tried with regression like 0.7xxxx, 0.000xxx..., no luck</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 710880,
          "author_name": "harshitsheoran",
          "author_url": "",
          "post_date": "01/05/2020 11:23:41",
          "content": "<h1>Case all 0</h1>\n\n<p>log_loss([\"REAL\", \"FAKE\"], [0, 0])\n-&gt; 17.269388197455342</p>\n\n<h1>Case all 1</h1>\n\n<p>log_loss([\"REAL\", \"FAKE\"], [1, 1])\n-&gt; 17.269388197455342</p>\n\n<h1>Case all right</h1>\n\n<p>log_loss([\"REAL\", \"FAKE\"], [1, 0])\n-&gt; 9.992007221626415e-16</p>\n\n<h1>Case all wrong</h1>\n\n<p>log_loss([\"REAL\", \"FAKE\"], [0, 1])\n-&gt; 34.53917619362578</p>\n\n<p>Thank You for your advise <a href=\"/lambrevam\">@lambrevam</a> but does not solve my problem...</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 711155,
      "author_name": "harshitsheoran",
      "author_url": "",
      "post_date": "01/05/2020 18:47:59",
      "content": "<p>Solved by infering not over the sample_sumbission videos but over the /testvideos/* files</p>\n\n<p>credit:- <a href=\"https://www.kaggle.com/c/deepfake-detection-challenge/discussion/123260\">https://www.kaggle.com/c/deepfake-detection-challenge/discussion/123260</a></p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 711330,
      "author_name": "marcuslin",
      "author_url": "",
      "post_date": "01/06/2020 00:10:07",
      "content": "<p>\"Solved by infering not over the sample_sumbission videos but over the /testvideos/* files\"   what do you mean?  I have the same issue,  I only use sample_submission file for generate finalsub,  thank in advance</p>",
      "votes": null,
      "replies": [
        {
          "id": 711393,
          "author_name": "harshitsheoran",
          "author_url": "",
          "post_date": "01/06/2020 03:02:43",
          "content": "<p>Hey <a href=\"/marcuslin\">@marcuslin</a>, do not use the sample sub either to iterate over videos or placing the output directly in sample submission...</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "710831": "Why all my subs are getting 17.26938 score? After submitting, the score gets calculated in a minute, does not take the time it takes to run the inference kernel, If someone knows the problem that would make my day, I am wasting my subs cuz of this thing.\n\nI request Kaggle Team to answer, there are a lot of candidates having this score, Thank You!\n\nEdit: About my sub:- It is around 200 0.0 and 200 1.0 on the public validation set, even if they are not right, that should still be around random, 0.69...",
    "710858": "Don't use 0.0 and 1.0.\nSubmissions are scored on log loss.\n\nhttps://www.kaggle.com/c/deepfake-detection-challenge/overview/evaluation",
    "710873": "I even tried with regression like 0.7xxxx, 0.000xxx..., no luck",
    "710880": "#Case all 0\nlog_loss([\"REAL\", \"FAKE\"], [0, 0])\n-&gt; 17.269388197455342\n#Case all 1\nlog_loss([\"REAL\", \"FAKE\"], [1, 1])\n-&gt; 17.269388197455342\n#Case all right\nlog_loss([\"REAL\", \"FAKE\"], [1, 0])\n-&gt; 9.992007221626415e-16\n#Case all wrong\nlog_loss([\"REAL\", \"FAKE\"], [0, 1])\n-&gt; 34.53917619362578\n\nThank You for your advise @lambrevam but does not solve my problem...",
    "711155": "Solved by infering not over the sample_sumbission videos but over the /testvideos/* files\n\ncredit:- https://www.kaggle.com/c/deepfake-detection-challenge/discussion/123260",
    "711330": "\"Solved by infering not over the sample_sumbission videos but over the /testvideos/* files\"   what do you mean?  I have the same issue,  I only use sample_submission file for generate finalsub,  thank in advance",
    "711393": "Hey @marcuslin, do not use the sample sub either to iterate over videos or placing the output directly in sample submission..."
  },
  "source": "meta"
}