{
  "id": 296024,
  "title": "Anyone succeeds with sequence_frame?",
  "url": "/competitions/tensorflow-great-barrier-reef/discussion/296024",
  "author_name": "",
  "post_date": "2021-12-19T11:35:33.252207300Z",
  "votes": 10,
  "comment_count": 7,
  "views": 0,
  "content": "<p>I couldn't utilize it so far 😔</p>",
  "messages": [
    {
      "id": "1623001",
      "postDate": "12/19/2021 11:35:33",
      "content": "<p>I couldn't utilize it so far 😔</p>",
      "rawMarkdown": "I couldn't utilize it so far 😔",
      "votes": null
    },
    {
      "id": "1623016",
      "postDate": "12/19/2021 11:50:18",
      "content": "<p>private test set does not provide sequence information, my inference code can only access private image once so pseudo labeling can't be apply for this comp too</p>",
      "rawMarkdown": "private test set does not provide sequence information, my inference code can only access private image once so pseudo labeling can't be apply for this comp too",
      "votes": null
    },
    {
      "id": "1623021",
      "postDate": "12/19/2021 11:58:27",
      "content": "<p>You mean the test.csv doesn't have <code>sequence_frame</code> in private? I could load test.csv in private, but haven't tried to use <code>sequence_frame</code> yet</p>",
      "rawMarkdown": "You mean the test.csv doesn't have `sequence_frame` in private? I could load test.csv in private, but haven't tried to use `sequence_frame` yet",
      "votes": null
    },
    {
      "id": "1623025",
      "postDate": "12/19/2021 12:05:26",
      "content": "<p>oh sorry, I missed test.csv :))</p>",
      "rawMarkdown": "oh sorry, I missed test.csv :))",
      "votes": null
    },
    {
      "id": "1623552",
      "postDate": "12/20/2021 02:20:45",
      "content": "<p>As a super beginner, still struggling with using groupKFolds based on Sequence.. I'm trying to apply what you did, but the performance is less than I expected.(still running ~fold3, but fold1 and fold2 seems way worse than the baseline using yolov5) Did you happen to use unlabeled images or other techniques? </p>",
      "rawMarkdown": "As a super beginner, still struggling with using groupKFolds based on Sequence.. I'm trying to apply what you did, but the performance is less than I expected.(still running ~fold3, but fold1 and fold2 seems way worse than the baseline using yolov5) Did you happen to use unlabeled images or other techniques?",
      "votes": null
    },
    {
      "id": "1623758",
      "postDate": "12/20/2021 08:04:41",
      "content": "<p>I’m not sure whether the test set images are fed in sequence order, so I’m also curious about how the sequence information is used😅</p>",
      "rawMarkdown": "I’m not sure whether the test set images are fed in sequence order, so I’m also curious about how the sequence information is used😅",
      "votes": null
    },
    {
      "id": "1623832",
      "postDate": "12/20/2021 10:05:42",
      "content": "<p>Fortunately they are :-)</p>\n<p>\"The API serves the images one by one, in order by video and frame number, as pixel arrays.\"</p>\n<p>See <a href=\"https://www.kaggle.com/c/tensorflow-great-barrier-reef/data#:~:text=The%20API%20serves%20the%20images%20one%20by%20one%2C%20in%20order%20by%20video%20and%20frame%20number%2C%20as%20pixel%20arrays.\" target=\"_blank\">here</a>.</p>",
      "rawMarkdown": "Fortunately they are :-)\n\n\"The API serves the images one by one, in order by video and frame number, as pixel arrays.\"\n\nSee [here](https://www.kaggle.com/c/tensorflow-great-barrier-reef/data#:~:text=The%20API%20serves%20the%20images%20one%20by%20one%2C%20in%20order%20by%20video%20and%20frame%20number%2C%20as%20pixel%20arrays.).",
      "votes": null
    },
    {
      "id": "1623852",
      "postDate": "12/20/2021 10:31:53",
      "content": "<p>Oh, thanks</p>",
      "rawMarkdown": "Oh, thanks",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1623016,
      "author_name": "ptran1203",
      "author_url": "",
      "post_date": "12/19/2021 11:50:18",
      "content": "<p>private test set does not provide sequence information, my inference code can only access private image once so pseudo labeling can't be apply for this comp too</p>",
      "votes": null,
      "replies": [
        {
          "id": 1623021,
          "author_name": "onodera",
          "author_url": "",
          "post_date": "12/19/2021 11:58:27",
          "content": "<p>You mean the test.csv doesn't have <code>sequence_frame</code> in private? I could load test.csv in private, but haven't tried to use <code>sequence_frame</code> yet</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1623025,
          "author_name": "ptran1203",
          "author_url": "",
          "post_date": "12/19/2021 12:05:26",
          "content": "<p>oh sorry, I missed test.csv :))</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1623552,
      "author_name": "joon9502",
      "author_url": "",
      "post_date": "12/20/2021 02:20:45",
      "content": "<p>As a super beginner, still struggling with using groupKFolds based on Sequence.. I'm trying to apply what you did, but the performance is less than I expected.(still running ~fold3, but fold1 and fold2 seems way worse than the baseline using yolov5) Did you happen to use unlabeled images or other techniques? </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1623758,
      "author_name": "klawensliu",
      "author_url": "",
      "post_date": "12/20/2021 08:04:41",
      "content": "<p>I’m not sure whether the test set images are fed in sequence order, so I’m also curious about how the sequence information is used😅</p>",
      "votes": null,
      "replies": [
        {
          "id": 1623832,
          "author_name": "hannes82",
          "author_url": "",
          "post_date": "12/20/2021 10:05:42",
          "content": "<p>Fortunately they are :-)</p>\n<p>\"The API serves the images one by one, in order by video and frame number, as pixel arrays.\"</p>\n<p>See <a href=\"https://www.kaggle.com/c/tensorflow-great-barrier-reef/data#:~:text=The%20API%20serves%20the%20images%20one%20by%20one%2C%20in%20order%20by%20video%20and%20frame%20number%2C%20as%20pixel%20arrays.\" target=\"_blank\">here</a>.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1623852,
          "author_name": "klawensliu",
          "author_url": "",
          "post_date": "12/20/2021 10:31:53",
          "content": "<p>Oh, thanks</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1623001": "I couldn't utilize it so far 😔",
    "1623016": "private test set does not provide sequence information, my inference code can only access private image once so pseudo labeling can't be apply for this comp too",
    "1623021": "You mean the test.csv doesn't have `sequence_frame` in private? I could load test.csv in private, but haven't tried to use `sequence_frame` yet",
    "1623025": "oh sorry, I missed test.csv :))",
    "1623552": "As a super beginner, still struggling with using groupKFolds based on Sequence.. I'm trying to apply what you did, but the performance is less than I expected.(still running ~fold3, but fold1 and fold2 seems way worse than the baseline using yolov5) Did you happen to use unlabeled images or other techniques?",
    "1623758": "I’m not sure whether the test set images are fed in sequence order, so I’m also curious about how the sequence information is used😅",
    "1623832": "Fortunately they are :-)\n\n\"The API serves the images one by one, in order by video and frame number, as pixel arrays.\"\n\nSee [here](https://www.kaggle.com/c/tensorflow-great-barrier-reef/data#:~:text=The%20API%20serves%20the%20images%20one%20by%20one%2C%20in%20order%20by%20video%20and%20frame%20number%2C%20as%20pixel%20arrays.).",
    "1623852": "Oh, thanks"
  },
  "source": "meta"
}