{
  "id": 149527,
  "title": "Is YOLOV3 effective ？",
  "url": "/competitions/global-wheat-detection/discussion/149527",
  "author_name": "",
  "post_date": "2020-05-08T16:08:36.806710100Z",
  "votes": null,
  "comment_count": 7,
  "views": 0,
  "content": "<p>Because YOLOV3 performs well in target detection, I want to know whether YOLOV3 is a good choice for this competition.</p>",
  "messages": [
    {
      "id": "838517",
      "postDate": "05/08/2020 16:08:36",
      "content": "<p>Because YOLOV3 performs well in target detection, I want to know whether YOLOV3 is a good choice for this competition.</p>",
      "rawMarkdown": "Because YOLOV3 performs well in target detection, I want to know whether YOLOV3 is a good choice for this competition.",
      "votes": null
    },
    {
      "id": "840741",
      "postDate": "05/10/2020 10:31:16",
      "content": "<p>Hello <a href=\"/gc1023\">@gc1023</a>,\nThe challenge organizers are talking in the intro about Yolov3 being used as one of the current detection methods, but being prone to bias by the training region. On their preprint (<a href=\"https://arxiv.org/ftp/arxiv/papers/2005/2005.02162.pdf\">https://arxiv.org/ftp/arxiv/papers/2005/2005.02162.pdf</a> ), the authors discussed Yolov3 as being used for semi-automatic labelling of their datasets, meaning that this method may be of interest to approach human annotation-level.\nSo for my part, i'm trying that !\nBest,\nJ.</p>",
      "rawMarkdown": "Hello @gc1023,\nThe challenge organizers are talking in the intro about Yolov3 being used as one of the current detection methods, but being prone to bias by the training region. On their preprint (https://arxiv.org/ftp/arxiv/papers/2005/2005.02162.pdf ), the authors discussed Yolov3 as being used for semi-automatic labelling of their datasets, meaning that this method may be of interest to approach human annotation-level.\nSo for my part, i'm trying that !\nBest,\nJ.",
      "votes": null
    },
    {
      "id": "840926",
      "postDate": "05/10/2020 13:49:30",
      "content": "<p>Thanks for your explanation, I will read this paper. Then, I will try. Not only yolov3 but also yolov4.</p>",
      "rawMarkdown": "Thanks for your explanation, I will read this paper. Then, I will try. Not only yolov3 but also yolov4.",
      "votes": null
    },
    {
      "id": "843300",
      "postDate": "05/12/2020 00:34:44",
      "content": "<p>I tried YOLOv4,  the best result so far is just 0.6 LB</p>",
      "rawMarkdown": "I tried YOLOv4,  the best result so far is just 0.6 LB",
      "votes": null
    },
    {
      "id": "843469",
      "postDate": "05/12/2020 04:06:57",
      "content": "<p>Thanks for sharing!</p>",
      "rawMarkdown": "Thanks for sharing!",
      "votes": null
    },
    {
      "id": "844011",
      "postDate": "05/12/2020 12:07:19",
      "content": "<p>I have published an inference notebook for YOLOv4 <a href=\"https://www.kaggle.com/nvnnghia/yolov4-inference\">here</a> check it out. The score is improved to 0.61</p>",
      "rawMarkdown": "I have published an inference notebook for YOLOv4 [here](https://www.kaggle.com/nvnnghia/yolov4-inference) check it out. The score is improved to 0.61",
      "votes": null
    },
    {
      "id": "844621",
      "postDate": "05/12/2020 17:57:29",
      "content": "<p>For my part, training of a out-of-the-box Yolov3 distro without any tweaking gave me 0,476.</p>",
      "rawMarkdown": "For my part, training of a out-of-the-box Yolov3 distro without any tweaking gave me 0,476.",
      "votes": null
    },
    {
      "id": "844967",
      "postDate": "05/13/2020 01:07:41",
      "content": "<p>I will learn your kernel. Thank you!</p>",
      "rawMarkdown": "I will learn your kernel. Thank you!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 840741,
      "author_name": "codename5281",
      "author_url": "",
      "post_date": "05/10/2020 10:31:16",
      "content": "<p>Hello <a href=\"/gc1023\">@gc1023</a>,\nThe challenge organizers are talking in the intro about Yolov3 being used as one of the current detection methods, but being prone to bias by the training region. On their preprint (<a href=\"https://arxiv.org/ftp/arxiv/papers/2005/2005.02162.pdf\">https://arxiv.org/ftp/arxiv/papers/2005/2005.02162.pdf</a> ), the authors discussed Yolov3 as being used for semi-automatic labelling of their datasets, meaning that this method may be of interest to approach human annotation-level.\nSo for my part, i'm trying that !\nBest,\nJ.</p>",
      "votes": null,
      "replies": [
        {
          "id": 840926,
          "author_name": "gc1023",
          "author_url": "",
          "post_date": "05/10/2020 13:49:30",
          "content": "<p>Thanks for your explanation, I will read this paper. Then, I will try. Not only yolov3 but also yolov4.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 843300,
      "author_name": "nvnnghia",
      "author_url": "",
      "post_date": "05/12/2020 00:34:44",
      "content": "<p>I tried YOLOv4,  the best result so far is just 0.6 LB</p>",
      "votes": null,
      "replies": [
        {
          "id": 843469,
          "author_name": "gc1023",
          "author_url": "",
          "post_date": "05/12/2020 04:06:57",
          "content": "<p>Thanks for sharing!</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 844011,
      "author_name": "nvnnghia",
      "author_url": "",
      "post_date": "05/12/2020 12:07:19",
      "content": "<p>I have published an inference notebook for YOLOv4 <a href=\"https://www.kaggle.com/nvnnghia/yolov4-inference\">here</a> check it out. The score is improved to 0.61</p>",
      "votes": null,
      "replies": [
        {
          "id": 844967,
          "author_name": "gc1023",
          "author_url": "",
          "post_date": "05/13/2020 01:07:41",
          "content": "<p>I will learn your kernel. Thank you!</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 844621,
      "author_name": "codename5281",
      "author_url": "",
      "post_date": "05/12/2020 17:57:29",
      "content": "<p>For my part, training of a out-of-the-box Yolov3 distro without any tweaking gave me 0,476.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "838517": "Because YOLOV3 performs well in target detection, I want to know whether YOLOV3 is a good choice for this competition.",
    "840741": "Hello @gc1023,\nThe challenge organizers are talking in the intro about Yolov3 being used as one of the current detection methods, but being prone to bias by the training region. On their preprint (https://arxiv.org/ftp/arxiv/papers/2005/2005.02162.pdf ), the authors discussed Yolov3 as being used for semi-automatic labelling of their datasets, meaning that this method may be of interest to approach human annotation-level.\nSo for my part, i'm trying that !\nBest,\nJ.",
    "840926": "Thanks for your explanation, I will read this paper. Then, I will try. Not only yolov3 but also yolov4.",
    "843300": "I tried YOLOv4,  the best result so far is just 0.6 LB",
    "843469": "Thanks for sharing!",
    "844011": "I have published an inference notebook for YOLOv4 [here](https://www.kaggle.com/nvnnghia/yolov4-inference) check it out. The score is improved to 0.61",
    "844621": "For my part, training of a out-of-the-box Yolov3 distro without any tweaking gave me 0,476.",
    "844967": "I will learn your kernel. Thank you!"
  },
  "source": "meta"
}