{
  "id": 229023,
  "title": "Datetime in metadata sometimes incorrect",
  "url": "/competitions/iwildcam2021-fgvc8/discussion/229023",
  "author_name": "",
  "post_date": "2021-03-27T20:11:25.638901700Z",
  "votes": 1,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I noticed that sometimes the datetime field of the metadata['images'] value is incorrect.</p>\n<p>An example is for seq_id 30251444-7d42-11eb-8fb5-0242ac1c0002 in which the datetime is the same for all images in the sequence. I don't think it's correct as even the display from the camera shows the time updating. Is it possible to fix this?</p>\n<p>Furthermore, for example in seq_id 302fd654-7d42-11eb-8fb5-0242ac1c0002 the datetime is listed the same in metadata, however in the camera it jumps more than 5 hours. I thought a sequence was supposed to be a short burst, and not a 5 hour+ long period?</p>",
  "messages": [
    {
      "id": "1254568",
      "postDate": "03/27/2021 20:11:25",
      "content": "<p>I noticed that sometimes the datetime field of the metadata['images'] value is incorrect.</p>\n<p>An example is for seq_id 30251444-7d42-11eb-8fb5-0242ac1c0002 in which the datetime is the same for all images in the sequence. I don't think it's correct as even the display from the camera shows the time updating. Is it possible to fix this?</p>\n<p>Furthermore, for example in seq_id 302fd654-7d42-11eb-8fb5-0242ac1c0002 the datetime is listed the same in metadata, however in the camera it jumps more than 5 hours. I thought a sequence was supposed to be a short burst, and not a 5 hour+ long period?</p>",
      "rawMarkdown": "I noticed that sometimes the datetime field of the metadata['images'] value is incorrect.\n\nAn example is for seq_id 30251444-7d42-11eb-8fb5-0242ac1c0002 in which the datetime is the same for all images in the sequence. I don't think it's correct as even the display from the camera shows the time updating. Is it possible to fix this?\n\nFurthermore, for example in seq_id 302fd654-7d42-11eb-8fb5-0242ac1c0002 the datetime is listed the same in metadata, however in the camera it jumps more than 5 hours. I thought a sequence was supposed to be a short burst, and not a 5 hour+ long period?",
      "votes": null
    },
    {
      "id": "1256111",
      "postDate": "03/29/2021 14:51:18",
      "content": "<p>We provide the metadata as it was provided to us by the Wildlife Conservation Society. We don't have access to the \"correct\" metadata in these error cases. Camera trap data is a very real-world challenge, and that means unfortunately that sometimes cameras malfunction or ecologists make mistakes.  I don't have corrections for these cases, but you are more than welcome to try to catch and correct these errors. Possibly OCR could be used to scrape the correct times from the image pixels? If you do find a solution, please share!</p>",
      "rawMarkdown": "We provide the metadata as it was provided to us by the Wildlife Conservation Society. We don't have access to the \"correct\" metadata in these error cases. Camera trap data is a very real-world challenge, and that means unfortunately that sometimes cameras malfunction or ecologists make mistakes.  I don't have corrections for these cases, but you are more than welcome to try to catch and correct these errors. Possibly OCR could be used to scrape the correct times from the image pixels? If you do find a solution, please share!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1256111,
      "author_name": "sbeery",
      "author_url": "",
      "post_date": "03/29/2021 14:51:18",
      "content": "<p>We provide the metadata as it was provided to us by the Wildlife Conservation Society. We don't have access to the \"correct\" metadata in these error cases. Camera trap data is a very real-world challenge, and that means unfortunately that sometimes cameras malfunction or ecologists make mistakes.  I don't have corrections for these cases, but you are more than welcome to try to catch and correct these errors. Possibly OCR could be used to scrape the correct times from the image pixels? If you do find a solution, please share!</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1254568": "I noticed that sometimes the datetime field of the metadata['images'] value is incorrect.\n\nAn example is for seq_id 30251444-7d42-11eb-8fb5-0242ac1c0002 in which the datetime is the same for all images in the sequence. I don't think it's correct as even the display from the camera shows the time updating. Is it possible to fix this?\n\nFurthermore, for example in seq_id 302fd654-7d42-11eb-8fb5-0242ac1c0002 the datetime is listed the same in metadata, however in the camera it jumps more than 5 hours. I thought a sequence was supposed to be a short burst, and not a 5 hour+ long period?",
    "1256111": "We provide the metadata as it was provided to us by the Wildlife Conservation Society. We don't have access to the \"correct\" metadata in these error cases. Camera trap data is a very real-world challenge, and that means unfortunately that sometimes cameras malfunction or ecologists make mistakes.  I don't have corrections for these cases, but you are more than welcome to try to catch and correct these errors. Possibly OCR could be used to scrape the correct times from the image pixels? If you do find a solution, please share!"
  },
  "source": "meta"
}