{
  "id": 303634,
  "title": "Yolov5 zero score?",
  "url": "/competitions/tensorflow-great-barrier-reef/discussion/303634",
  "author_name": "",
  "post_date": "2022-01-28T14:21:36.522672Z",
  "votes": 1,
  "comment_count": 6,
  "views": 0,
  "content": "<p>I've submitted two yolov5 models, but my submission were zero-scored. What can be the possible reason for that?</p>",
  "messages": [
    {
      "id": "1667390",
      "postDate": "01/28/2022 14:21:36",
      "content": "<p>I've submitted two yolov5 models, but my submission were zero-scored. What can be the possible reason for that?</p>",
      "rawMarkdown": "I've submitted two yolov5 models, but my submission were zero-scored. What can be the possible reason for that?",
      "votes": null
    },
    {
      "id": "1667540",
      "postDate": "01/28/2022 17:40:47",
      "content": "<p>Something strange with boxes coordinates (not proper format) or … with score …. I guess …</p>",
      "rawMarkdown": "Something strange with boxes coordinates (not proper format) or … with score …. I guess …",
      "votes": null
    },
    {
      "id": "1667553",
      "postDate": "01/28/2022 17:56:08",
      "content": "<p>Most probably first option is my case. I'll double check it. Thanks!</p>",
      "rawMarkdown": "Most probably first option is my case. I'll double check it. Thanks!",
      "votes": null
    },
    {
      "id": "1667564",
      "postDate": "01/28/2022 18:02:40",
      "content": "<p>The most common problem is to exchange class with score :) Hope you find problem and progress. 👍</p>",
      "rawMarkdown": "The most common problem is to exchange class with score :) Hope you find problem and progress. 👍",
      "votes": null
    },
    {
      "id": "1667588",
      "postDate": "01/28/2022 18:46:16",
      "content": "<p>I've made a video with my model prediction with <a href=\"https://www.kaggle.com/alexchwong/yolov5-is-all-you-need-make-compact-videos\" target=\"_blank\">this notebook</a>, and it seems that my model gives MANY false detections. Maybe a reason for that is a small training size (1280).</p>",
      "rawMarkdown": "I've made a video with my model prediction with [this notebook](https://www.kaggle.com/alexchwong/yolov5-is-all-you-need-make-compact-videos), and it seems that my model gives MANY false detections. Maybe a reason for that is a small training size (1280).",
      "votes": null
    },
    {
      "id": "1667611",
      "postDate": "01/28/2022 19:11:45",
      "content": "<p>Check that your model is being tested on the same color channel format as the training data. Try changing the order of the pixel array by swapping the red and blue color channels.</p>",
      "rawMarkdown": "Check that your model is being tested on the same color channel format as the training data. Try changing the order of the pixel array by swapping the red and blue color channels.",
      "votes": null
    },
    {
      "id": "1667620",
      "postDate": "01/28/2022 19:20:25",
      "content": "<p>Thanks, I'll check it.</p>",
      "rawMarkdown": "Thanks, I'll check it.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1667540,
      "author_name": "remekkinas",
      "author_url": "",
      "post_date": "01/28/2022 17:40:47",
      "content": "<p>Something strange with boxes coordinates (not proper format) or … with score …. I guess …</p>",
      "votes": null,
      "replies": [
        {
          "id": 1667553,
          "author_name": "atamazian",
          "author_url": "",
          "post_date": "01/28/2022 17:56:08",
          "content": "<p>Most probably first option is my case. I'll double check it. Thanks!</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1667564,
          "author_name": "remekkinas",
          "author_url": "",
          "post_date": "01/28/2022 18:02:40",
          "content": "<p>The most common problem is to exchange class with score :) Hope you find problem and progress. 👍</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1667588,
          "author_name": "atamazian",
          "author_url": "",
          "post_date": "01/28/2022 18:46:16",
          "content": "<p>I've made a video with my model prediction with <a href=\"https://www.kaggle.com/alexchwong/yolov5-is-all-you-need-make-compact-videos\" target=\"_blank\">this notebook</a>, and it seems that my model gives MANY false detections. Maybe a reason for that is a small training size (1280).</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1667611,
      "author_name": "maxvandijck",
      "author_url": "",
      "post_date": "01/28/2022 19:11:45",
      "content": "<p>Check that your model is being tested on the same color channel format as the training data. Try changing the order of the pixel array by swapping the red and blue color channels.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1667620,
          "author_name": "atamazian",
          "author_url": "",
          "post_date": "01/28/2022 19:20:25",
          "content": "<p>Thanks, I'll check it.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1667390": "I've submitted two yolov5 models, but my submission were zero-scored. What can be the possible reason for that?",
    "1667540": "Something strange with boxes coordinates (not proper format) or … with score …. I guess …",
    "1667553": "Most probably first option is my case. I'll double check it. Thanks!",
    "1667564": "The most common problem is to exchange class with score :) Hope you find problem and progress. 👍",
    "1667588": "I've made a video with my model prediction with [this notebook](https://www.kaggle.com/alexchwong/yolov5-is-all-you-need-make-compact-videos), and it seems that my model gives MANY false detections. Maybe a reason for that is a small training size (1280).",
    "1667611": "Check that your model is being tested on the same color channel format as the training data. Try changing the order of the pixel array by swapping the red and blue color channels.",
    "1667620": "Thanks, I'll check it."
  },
  "source": "meta"
}