{
  "id": 249296,
  "title": "flexible-yolov5 : Reorganize model structure, such as backbone, neck, head, can modify the network flexibly and conveniently",
  "url": "/competitions/siim-covid19-detection/discussion/249296",
  "author_name": "",
  "post_date": "2021-06-27T16:39:46.537570Z",
  "votes": 24,
  "comment_count": 5,
  "views": 0,
  "content": "<p><a href=\"https://github.com/yl305237731/flexible-yolov5\" target=\"_blank\">https://github.com/yl305237731/flexible-yolov5</a></p>\n<p>Features</p>\n<ol>\n<li>mobilenetV3-small, mobilenetV3-large</li>\n<li>shufflenet_v2_x0_5, shufflenet_v2_x1_0, shufflenet_v2_x1_5, shufflenet_v2_x2_0</li>\n<li>yolov5s, yolov5m, yolov5l, yolov5x, yolov5transformer</li>\n<li>resnet18, resnet50, resnet34, resnet101, resnet152</li>\n<li>efficientnet_b0 - efficientnet_b8, efficientnet_l2</li>\n<li>hrnet 18,32,48</li>\n<li>CBAM, SE</li>\n<li>DCN (mixed precision training not support, if you want use dcn, please close amp in line 292 of scripts/train.py)</li>\n<li>coord conv</li>\n<li>drop_block</li>\n</ol>",
  "messages": [
    {
      "id": "1367374",
      "postDate": "06/27/2021 16:39:46",
      "content": "<p><a href=\"https://github.com/yl305237731/flexible-yolov5\" target=\"_blank\">https://github.com/yl305237731/flexible-yolov5</a></p>\n<p>Features</p>\n<ol>\n<li>mobilenetV3-small, mobilenetV3-large</li>\n<li>shufflenet_v2_x0_5, shufflenet_v2_x1_0, shufflenet_v2_x1_5, shufflenet_v2_x2_0</li>\n<li>yolov5s, yolov5m, yolov5l, yolov5x, yolov5transformer</li>\n<li>resnet18, resnet50, resnet34, resnet101, resnet152</li>\n<li>efficientnet_b0 - efficientnet_b8, efficientnet_l2</li>\n<li>hrnet 18,32,48</li>\n<li>CBAM, SE</li>\n<li>DCN (mixed precision training not support, if you want use dcn, please close amp in line 292 of scripts/train.py)</li>\n<li>coord conv</li>\n<li>drop_block</li>\n</ol>",
      "rawMarkdown": "https://github.com/yl305237731/flexible-yolov5\n\nFeatures\n1. mobilenetV3-small, mobilenetV3-large\n2. shufflenet_v2_x0_5, shufflenet_v2_x1_0, shufflenet_v2_x1_5, shufflenet_v2_x2_0\n3. yolov5s, yolov5m, yolov5l, yolov5x, yolov5transformer\n4. resnet18, resnet50, resnet34, resnet101, resnet152\n5. efficientnet_b0 - efficientnet_b8, efficientnet_l2\n6. hrnet 18,32,48\n7. CBAM, SE\n8. DCN (mixed precision training not support, if you want use dcn, please close amp in line 292 of scripts/train.py)\n9. coord conv\n10. drop_block",
      "votes": null
    },
    {
      "id": "1393854",
      "postDate": "07/20/2021 01:57:19",
      "content": "<p>If you have sample notebook or any stater notebook of this, then please share it. I tried to use it but it always throw an error.</p>",
      "rawMarkdown": "If you have sample notebook or any stater notebook of this, then please share it. I tried to use it but it always throw an error.",
      "votes": null
    },
    {
      "id": "1398471",
      "postDate": "07/24/2021 07:50:19",
      "content": "<p>I managed to use it for training, only on a local computer, because, for some reason, I can't download weights if I train in klaggle. resnet 101, efnb6 showed a rather weak result</p>",
      "rawMarkdown": "I managed to use it for training, only on a local computer, because, for some reason, I can't download weights if I train in klaggle. resnet 101, efnb6 showed a rather weak result",
      "votes": null
    },
    {
      "id": "1398721",
      "postDate": "07/24/2021 12:34:41",
      "content": "<p>Can you please share the scripts or notebook? If you do not want to share it, it's okay.</p>",
      "rawMarkdown": "Can you please share the scripts or notebook? If you do not want to share it, it's okay.",
      "votes": null
    },
    {
      "id": "1398742",
      "postDate": "07/24/2021 13:00:23",
      "content": "<p>he's gone. copy the library to your working directory. install dependencies and start a workout<br>\nbefore doing this, prepare the dataset<br>\nthen change data.yaml and customize  confiпs/model_XXX.yaml</p>",
      "rawMarkdown": "he's gone. copy the library to your working directory. install dependencies and start a workout\nbefore doing this, prepare the dataset\nthen change data.yaml and customize  confiпs/model_XXX.yaml",
      "votes": null
    },
    {
      "id": "1398745",
      "postDate": "07/24/2021 13:03:59",
      "content": "<p>shutil.copytree('/kaggle/input/flexible-yolov5/flexible-yolov5', '/kaggle/working/yolov5')<br>\nos.chdir('/kaggle/working/yolov5')<br>\n!pip install -r requirements.txt<br>\n!pip install addict<br>\n!WANDB_MODE=\"dryrun\" python scripts/train.py --batch 4 --epochs 30 --data configs/data.yaml --cfg configs/model_efficientnet.yaml</p>",
      "rawMarkdown": "shutil.copytree('/kaggle/input/flexible-yolov5/flexible-yolov5', '/kaggle/working/yolov5')\nos.chdir('/kaggle/working/yolov5')\n!pip install -r requirements.txt\n!pip install addict\n!WANDB_MODE=\"dryrun\" python scripts/train.py --batch 4 --epochs 30 --data configs/data.yaml --cfg configs/model_efficientnet.yaml",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1393854,
      "author_name": "vatsalmavani",
      "author_url": "",
      "post_date": "07/20/2021 01:57:19",
      "content": "<p>If you have sample notebook or any stater notebook of this, then please share it. I tried to use it but it always throw an error.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1398471,
          "author_name": "zaakciiru",
          "author_url": "",
          "post_date": "07/24/2021 07:50:19",
          "content": "<p>I managed to use it for training, only on a local computer, because, for some reason, I can't download weights if I train in klaggle. resnet 101, efnb6 showed a rather weak result</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1398721,
          "author_name": "vatsalmavani",
          "author_url": "",
          "post_date": "07/24/2021 12:34:41",
          "content": "<p>Can you please share the scripts or notebook? If you do not want to share it, it's okay.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1398742,
          "author_name": "zaakciiru",
          "author_url": "",
          "post_date": "07/24/2021 13:00:23",
          "content": "<p>he's gone. copy the library to your working directory. install dependencies and start a workout<br>\nbefore doing this, prepare the dataset<br>\nthen change data.yaml and customize  confiпs/model_XXX.yaml</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1398745,
          "author_name": "zaakciiru",
          "author_url": "",
          "post_date": "07/24/2021 13:03:59",
          "content": "<p>shutil.copytree('/kaggle/input/flexible-yolov5/flexible-yolov5', '/kaggle/working/yolov5')<br>\nos.chdir('/kaggle/working/yolov5')<br>\n!pip install -r requirements.txt<br>\n!pip install addict<br>\n!WANDB_MODE=\"dryrun\" python scripts/train.py --batch 4 --epochs 30 --data configs/data.yaml --cfg configs/model_efficientnet.yaml</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1367374": "https://github.com/yl305237731/flexible-yolov5\n\nFeatures\n1. mobilenetV3-small, mobilenetV3-large\n2. shufflenet_v2_x0_5, shufflenet_v2_x1_0, shufflenet_v2_x1_5, shufflenet_v2_x2_0\n3. yolov5s, yolov5m, yolov5l, yolov5x, yolov5transformer\n4. resnet18, resnet50, resnet34, resnet101, resnet152\n5. efficientnet_b0 - efficientnet_b8, efficientnet_l2\n6. hrnet 18,32,48\n7. CBAM, SE\n8. DCN (mixed precision training not support, if you want use dcn, please close amp in line 292 of scripts/train.py)\n9. coord conv\n10. drop_block",
    "1393854": "If you have sample notebook or any stater notebook of this, then please share it. I tried to use it but it always throw an error.",
    "1398471": "I managed to use it for training, only on a local computer, because, for some reason, I can't download weights if I train in klaggle. resnet 101, efnb6 showed a rather weak result",
    "1398721": "Can you please share the scripts or notebook? If you do not want to share it, it's okay.",
    "1398742": "he's gone. copy the library to your working directory. install dependencies and start a workout\nbefore doing this, prepare the dataset\nthen change data.yaml and customize  confiпs/model_XXX.yaml",
    "1398745": "shutil.copytree('/kaggle/input/flexible-yolov5/flexible-yolov5', '/kaggle/working/yolov5')\nos.chdir('/kaggle/working/yolov5')\n!pip install -r requirements.txt\n!pip install addict\n!WANDB_MODE=\"dryrun\" python scripts/train.py --batch 4 --epochs 30 --data configs/data.yaml --cfg configs/model_efficientnet.yaml"
  },
  "source": "meta"
}