{
  "id": 220524,
  "title": "Learning from this competition",
  "url": "/competitions/cassava-leaf-disease-classification/discussion/220524",
  "author_name": "Trushant Kalyanpur",
  "post_date": "2021-02-18T17:02:54.317000",
  "votes": 4,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Although I had a late start to this competition, if I can take something away from this, is that it taught me the importance of different types of augmentations like Cutmix/Fmix/ tweaking Albumentation augmentations depending on the noise in the dataset and how the choice of learning rate schedulers and picking the right loss function matters for different use cases.</p>\n<p>I now have a stronger Pytorch based template which I am sure will help me in future competitions. </p>\n<p>I am sure everyone has something that this unique competition has taught them. </p>",
  "messages": [
    {
      "id": 1209073,
      "postDate": "2021-02-18T17:02:54.317Z",
      "content": "<p>Although I had a late start to this competition, if I can take something away from this, is that it taught me the importance of different types of augmentations like Cutmix/Fmix/ tweaking Albumentation augmentations depending on the noise in the dataset and how the choice of learning rate schedulers and picking the right loss function matters for different use cases.</p>\n<p>I now have a stronger Pytorch based template which I am sure will help me in future competitions. </p>\n<p>I am sure everyone has something that this unique competition has taught them. </p>",
      "rawMarkdown": "Although I had a late start to this competition, if I can take something away from this, is that it taught me the importance of different types of augmentations like Cutmix/Fmix/ tweaking Albumentation augmentations depending on the noise in the dataset and how the choice of learning rate schedulers and picking the right loss function matters for different use cases.\n\nI now have a stronger Pytorch based template which I am sure will help me in future competitions. \n\nI am sure everyone has something that this unique competition has taught them. ",
      "votes": 4
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1209073": "Although I had a late start to this competition, if I can take something away from this, is that it taught me the importance of different types of augmentations like Cutmix/Fmix/ tweaking Albumentation augmentations depending on the noise in the dataset and how the choice of learning rate schedulers and picking the right loss function matters for different use cases.\n\nI now have a stronger Pytorch based template which I am sure will help me in future competitions. \n\nI am sure everyone has something that this unique competition has taught them. "
  }
}