{
  "id": 311838,
  "title": "Summary of previous competition solutions (continued)",
  "url": "/competitions/snakeclef2022/discussion/311838",
  "author_name": "",
  "post_date": "2022-03-09T03:13:15.676066400Z",
  "votes": 2,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hi all,</p>\n<p>I am starting this thread to post solutions of previous SnakeCLEF competitions. Please feel free to add on to this thread!</p>\n<ul>\n<li><p>SnakeCLEF 2020</p>\n<ul>\n<li>Classes: 783 species.</li>\n<li>Number of training images: 245,185 images.</li>\n<li>1st solution.<ul>\n<li>ResNet50-V2 pretrained on ImageNet-21K. F1: 0.625</li>\n<li>Included geolocations. </li></ul></li></ul></li>\n<li><p>SnakeCLEF 2021</p>\n<ul>\n<li>Classes: 772 species</li>\n<li>Number of training images: 414,424 images. </li>\n<li>Additional geolocation (country and continent).</li>\n<li>5th solution.<br>\n-ResNet-18. F1: 0.74, F1-country: 0.78, Accuracy: 0.86<br>\n-Frozen the first three layers as is the pretrained state with ImageNet-1K.<br>\n-Dropout: 0.5 within ResNet.<br>\n-Trained for 14 Epochs.<br>\n-Included geolocations in training.<br>\n-Did not address class imbalance. </li></ul></li>\n</ul>\n<p>Reference:<br>\n<a href=\"http://ceur-ws.org/Vol-2936/paper-130.pdf\" target=\"_blank\">http://ceur-ws.org/Vol-2936/paper-130.pdf</a></p>",
  "messages": [
    {
      "id": "1716449",
      "postDate": "03/09/2022 03:13:15",
      "content": "<p>Hi all,</p>\n<p>I am starting this thread to post solutions of previous SnakeCLEF competitions. Please feel free to add on to this thread!</p>\n<ul>\n<li><p>SnakeCLEF 2020</p>\n<ul>\n<li>Classes: 783 species.</li>\n<li>Number of training images: 245,185 images.</li>\n<li>1st solution.<ul>\n<li>ResNet50-V2 pretrained on ImageNet-21K. F1: 0.625</li>\n<li>Included geolocations. </li></ul></li></ul></li>\n<li><p>SnakeCLEF 2021</p>\n<ul>\n<li>Classes: 772 species</li>\n<li>Number of training images: 414,424 images. </li>\n<li>Additional geolocation (country and continent).</li>\n<li>5th solution.<br>\n-ResNet-18. F1: 0.74, F1-country: 0.78, Accuracy: 0.86<br>\n-Frozen the first three layers as is the pretrained state with ImageNet-1K.<br>\n-Dropout: 0.5 within ResNet.<br>\n-Trained for 14 Epochs.<br>\n-Included geolocations in training.<br>\n-Did not address class imbalance. </li></ul></li>\n</ul>\n<p>Reference:<br>\n<a href=\"http://ceur-ws.org/Vol-2936/paper-130.pdf\" target=\"_blank\">http://ceur-ws.org/Vol-2936/paper-130.pdf</a></p>",
      "rawMarkdown": "Hi all,\n\nI am starting this thread to post solutions of previous SnakeCLEF competitions. Please feel free to add on to this thread!\n\n\n- SnakeCLEF 2020\n  - Classes: 783 species.\n  - Number of training images: 245,185 images.\n  - 1st solution.\n     - ResNet50-V2 pretrained on ImageNet-21K. F1: 0.625\n     - Included geolocations. \n\n- SnakeCLEF 2021\n  - Classes: 772 species\n  - Number of training images: 414,424 images. \n  - Additional geolocation (country and continent).\n  - 5th solution.\n      -ResNet-18. F1: 0.74, F1-country: 0.78, Accuracy: 0.86\n      -Frozen the first three layers as is the pretrained state with ImageNet-1K.\n      -Dropout: 0.5 within ResNet.\n      -Trained for 14 Epochs.\n      -Included geolocations in training.\n      -Did not address class imbalance. \n\n \n\n\nReference:\nhttp://ceur-ws.org/Vol-2936/paper-130.pdf",
      "votes": null
    },
    {
      "id": "1717933",
      "postDate": "03/10/2022 11:05:24",
      "content": "<p>Hi John,</p>\n<p>The official results and short description (Including winning submission) with referencing could be found in this paper --&gt; <a href=\"http://ceur-ws.org/Vol-2936/paper-125.pdf\" target=\"_blank\">http://ceur-ws.org/Vol-2936/paper-125.pdf</a></p>\n<p>Best,<br>\nLukas</p>",
      "rawMarkdown": "Hi John,\n\nThe official results and short description (Including winning submission) with referencing could be found in this paper --> http://ceur-ws.org/Vol-2936/paper-125.pdf\n\nBest,\nLukas",
      "votes": null
    },
    {
      "id": "1814982",
      "postDate": "06/08/2022 14:35:05",
      "content": "<p>Please refer to my topic in discussion, I need your help, your upvote, thanks. <a href=\"https://www.kaggle.com/parkjohnychae\" target=\"_blank\">@parkjohnychae</a> </p>",
      "rawMarkdown": "Please refer to my topic in discussion, I need your help, your upvote, thanks. @parkjohnychae",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1717933,
      "author_name": "picekl",
      "author_url": "",
      "post_date": "03/10/2022 11:05:24",
      "content": "<p>Hi John,</p>\n<p>The official results and short description (Including winning submission) with referencing could be found in this paper --&gt; <a href=\"http://ceur-ws.org/Vol-2936/paper-125.pdf\" target=\"_blank\">http://ceur-ws.org/Vol-2936/paper-125.pdf</a></p>\n<p>Best,<br>\nLukas</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1814982,
      "author_name": "abuduori",
      "author_url": "",
      "post_date": "06/08/2022 14:35:05",
      "content": "<p>Please refer to my topic in discussion, I need your help, your upvote, thanks. <a href=\"https://www.kaggle.com/parkjohnychae\" target=\"_blank\">@parkjohnychae</a> </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1716449": "Hi all,\n\nI am starting this thread to post solutions of previous SnakeCLEF competitions. Please feel free to add on to this thread!\n\n\n- SnakeCLEF 2020\n  - Classes: 783 species.\n  - Number of training images: 245,185 images.\n  - 1st solution.\n     - ResNet50-V2 pretrained on ImageNet-21K. F1: 0.625\n     - Included geolocations. \n\n- SnakeCLEF 2021\n  - Classes: 772 species\n  - Number of training images: 414,424 images. \n  - Additional geolocation (country and continent).\n  - 5th solution.\n      -ResNet-18. F1: 0.74, F1-country: 0.78, Accuracy: 0.86\n      -Frozen the first three layers as is the pretrained state with ImageNet-1K.\n      -Dropout: 0.5 within ResNet.\n      -Trained for 14 Epochs.\n      -Included geolocations in training.\n      -Did not address class imbalance. \n\n \n\n\nReference:\nhttp://ceur-ws.org/Vol-2936/paper-130.pdf",
    "1717933": "Hi John,\n\nThe official results and short description (Including winning submission) with referencing could be found in this paper --> http://ceur-ws.org/Vol-2936/paper-125.pdf\n\nBest,\nLukas",
    "1814982": "Please refer to my topic in discussion, I need your help, your upvote, thanks. @parkjohnychae"
  },
  "source": "meta"
}