{
  "id": 55927,
  "title": "Submission format is unclear to me, getting only 0 scores",
  "url": "/competitions/cvpr-2018-autonomous-driving/discussion/55927",
  "author_name": "",
  "post_date": "2018-05-03T07:58:50.993528900Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>I tried many things to create a working submission.\n- I used this kernel <a href=\"https://www.kaggle.com/kmader/opencv-hog-submission\">https://www.kaggle.com/kmader/opencv-hog-submission</a>\n   which produce reasonable results for persons.\n   Got 0 score\n-  I tried to predict whole images as 33 label. That supose to get some low score.\n   The format is  simple:\n   PixelCount is: with*height\n   PixelEncoding is: \"0, width*height|\". \n  I got an error.\n- All the rest I tried, timed out or produced zero.\nMaybe the organizers can set up a small test submission for one or few images. Just to check the format and size limits. \nOr produce a code to create submission from predicted labels.\nOr release some example with submission larger then zero.\nAny advice on the submission format will be appreciated.</p>",
  "messages": [
    {
      "id": "322570",
      "postDate": "05/03/2018 07:58:50",
      "content": "<p>I tried many things to create a working submission.\n- I used this kernel <a href=\"https://www.kaggle.com/kmader/opencv-hog-submission\">https://www.kaggle.com/kmader/opencv-hog-submission</a>\n   which produce reasonable results for persons.\n   Got 0 score\n-  I tried to predict whole images as 33 label. That supose to get some low score.\n   The format is  simple:\n   PixelCount is: with*height\n   PixelEncoding is: \"0, width*height|\". \n  I got an error.\n- All the rest I tried, timed out or produced zero.\nMaybe the organizers can set up a small test submission for one or few images. Just to check the format and size limits. \nOr produce a code to create submission from predicted labels.\nOr release some example with submission larger then zero.\nAny advice on the submission format will be appreciated.</p>",
      "rawMarkdown": "I tried many things to create a working submission.\n- I used this kernel https://www.kaggle.com/kmader/opencv-hog-submission\n   which produce reasonable results for persons.\n   Got 0 score\n-  I tried to predict whole images as 33 label. That supose to get some low score.\n   The format is  simple:\n   PixelCount is: with*height\n   PixelEncoding is: \"0, width*height|\". \n  I got an error.\n- All the rest I tried, timed out or produced zero.\nMaybe the organizers can set up a small test submission for one or few images. Just to check the format and size limits. \nOr produce a code to create submission from predicted labels.\nOr release some example with submission larger then zero.\nAny advice on the submission format will be appreciated.",
      "votes": null
    },
    {
      "id": "324061",
      "postDate": "05/07/2018 03:38:24",
      "content": "<p>We have uploaded the codes that convert label images to CSV format. Here are some suggestions: 1. Check if your submissions are row-major order and zero-indexed; 2. If you try to predict whole image as 33, you will get 0 score or an error.</p>",
      "rawMarkdown": "We have uploaded the codes that convert label images to CSV format. Here are some suggestions: 1. Check if your submissions are row-major order and zero-indexed; 2. If you try to predict whole image as 33, you will get 0 score or an error.",
      "votes": null
    },
    {
      "id": "337653",
      "postDate": "06/03/2018 12:24:28",
      "content": "<p>I have similar problem. submission file is about 40 mb, it is in row-major order and zero-indexed and all confidence values are greater than 0.5 but score is 0.</p>",
      "rawMarkdown": "I have similar problem. submission file is about 40 mb, it is in row-major order and zero-indexed and all confidence values are greater than 0.5 but score is 0.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 324061,
      "author_name": "huangxinyu01",
      "author_url": "",
      "post_date": "05/07/2018 03:38:24",
      "content": "<p>We have uploaded the codes that convert label images to CSV format. Here are some suggestions: 1. Check if your submissions are row-major order and zero-indexed; 2. If you try to predict whole image as 33, you will get 0 score or an error.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 337653,
      "author_name": "yzokulu",
      "author_url": "",
      "post_date": "06/03/2018 12:24:28",
      "content": "<p>I have similar problem. submission file is about 40 mb, it is in row-major order and zero-indexed and all confidence values are greater than 0.5 but score is 0.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "322570": "I tried many things to create a working submission.\n- I used this kernel https://www.kaggle.com/kmader/opencv-hog-submission\n   which produce reasonable results for persons.\n   Got 0 score\n-  I tried to predict whole images as 33 label. That supose to get some low score.\n   The format is  simple:\n   PixelCount is: with*height\n   PixelEncoding is: \"0, width*height|\". \n  I got an error.\n- All the rest I tried, timed out or produced zero.\nMaybe the organizers can set up a small test submission for one or few images. Just to check the format and size limits. \nOr produce a code to create submission from predicted labels.\nOr release some example with submission larger then zero.\nAny advice on the submission format will be appreciated.",
    "324061": "We have uploaded the codes that convert label images to CSV format. Here are some suggestions: 1. Check if your submissions are row-major order and zero-indexed; 2. If you try to predict whole image as 33, you will get 0 score or an error.",
    "337653": "I have similar problem. submission file is about 40 mb, it is in row-major order and zero-indexed and all confidence values are greater than 0.5 but score is 0."
  },
  "source": "meta"
}