{
  "id": 233032,
  "title": "Model Weights for Resnet18 and Resnet50 (LB: 0.24)",
  "url": "/competitions/herbarium-2021-fgvc8/discussion/233032",
  "author_name": "Saurav Maheshkar ☕️",
  "post_date": "2021-04-16T18:44:59.142000",
  "votes": 3,
  "comment_count": 0,
  "views": 0,
  "content": "<p>I shared the model weights for Resnet18 and Resnet50 trained for 4 epochs. The Model Weights can be found in this dataset: <a href=\"https://www.kaggle.com/sauravmaheshkar/herbarium-2021-pytorch-weights\" target=\"_blank\">Herbarium🌿2021: Pytorch 🔥Model Weights</a>. </p>\n<p>I used the Resnet50 weights to obtain a score of ~0.24 on the Leaderboard. I followed the same procedure as in this notebook: <a href=\"https://www.kaggle.com/sauravmaheshkar/herbarium-2021-resnet18-inference\" target=\"_blank\">Herbarium 2021: Resnet18 Inference</a></p>\n<p>I didn't experiment with a bottleneck layer for these models. Although I'd recommend a higher model for that perhaps <code>resnet101</code> or <code>resnext101_32x8d</code></p>",
  "messages": [
    {
      "id": 1275844,
      "postDate": "2021-04-16T18:44:59.143Z",
      "content": "<p>I shared the model weights for Resnet18 and Resnet50 trained for 4 epochs. The Model Weights can be found in this dataset: <a href=\"https://www.kaggle.com/sauravmaheshkar/herbarium-2021-pytorch-weights\" target=\"_blank\">Herbarium🌿2021: Pytorch 🔥Model Weights</a>. </p>\n<p>I used the Resnet50 weights to obtain a score of ~0.24 on the Leaderboard. I followed the same procedure as in this notebook: <a href=\"https://www.kaggle.com/sauravmaheshkar/herbarium-2021-resnet18-inference\" target=\"_blank\">Herbarium 2021: Resnet18 Inference</a></p>\n<p>I didn't experiment with a bottleneck layer for these models. Although I'd recommend a higher model for that perhaps <code>resnet101</code> or <code>resnext101_32x8d</code></p>",
      "rawMarkdown": "I shared the model weights for Resnet18 and Resnet50 trained for 4 epochs. The Model Weights can be found in this dataset: [Herbarium🌿2021: Pytorch 🔥Model Weights](https://www.kaggle.com/sauravmaheshkar/herbarium-2021-pytorch-weights). \n\nI used the Resnet50 weights to obtain a score of ~0.24 on the Leaderboard. I followed the same procedure as in this notebook: [Herbarium 2021: Resnet18 Inference](https://www.kaggle.com/sauravmaheshkar/herbarium-2021-resnet18-inference)\n\n\nI didn't experiment with a bottleneck layer for these models. Although I'd recommend a higher model for that perhaps `resnet101` or `resnext101_32x8d`\n\n\n",
      "votes": 3
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1275844": "I shared the model weights for Resnet18 and Resnet50 trained for 4 epochs. The Model Weights can be found in this dataset: [Herbarium🌿2021: Pytorch 🔥Model Weights](https://www.kaggle.com/sauravmaheshkar/herbarium-2021-pytorch-weights). \n\nI used the Resnet50 weights to obtain a score of ~0.24 on the Leaderboard. I followed the same procedure as in this notebook: [Herbarium 2021: Resnet18 Inference](https://www.kaggle.com/sauravmaheshkar/herbarium-2021-resnet18-inference)\n\n\nI didn't experiment with a bottleneck layer for these models. Although I'd recommend a higher model for that perhaps `resnet101` or `resnext101_32x8d`\n\n\n"
  }
}