{
  "id": 40508,
  "title": "Considerations for fine-grained image classification?",
  "url": "/competitions/dog-breed-identification/discussion/40508",
  "author_name": "Pedro Nascimento",
  "post_date": "2017-10-03T17:02:34.130000",
  "votes": 1,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hello all</p>\n\n<p>This competition it seems is a more accessible version of ILSVRC, but dealing only with fine grained classification. Given that this competition's data is a subset of Imagenet, there is no doubt that the models trained on it would have enough capacity to fit the data, but I was wondering, is there any architecture/model specialization to deal specifically with the problem of fine-grained classes ?</p>\n\n<p>Cheers</p>",
  "messages": [
    {
      "id": 227075,
      "postDate": "2017-10-03T17:02:34.130Z",
      "content": "<p>Hello all</p>\n\n<p>This competition it seems is a more accessible version of ILSVRC, but dealing only with fine grained classification. Given that this competition's data is a subset of Imagenet, there is no doubt that the models trained on it would have enough capacity to fit the data, but I was wondering, is there any architecture/model specialization to deal specifically with the problem of fine-grained classes ?</p>\n\n<p>Cheers</p>",
      "rawMarkdown": "Hello all\n\nThis competition it seems is a more accessible version of ILSVRC, but dealing only with fine grained classification. Given that this competition's data is a subset of Imagenet, there is no doubt that the models trained on it would have enough capacity to fit the data, but I was wondering, is there any architecture/model specialization to deal specifically with the problem of fine-grained classes ?\n\nCheers",
      "votes": 1
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "227075": "Hello all\n\nThis competition it seems is a more accessible version of ILSVRC, but dealing only with fine grained classification. Given that this competition's data is a subset of Imagenet, there is no doubt that the models trained on it would have enough capacity to fit the data, but I was wondering, is there any architecture/model specialization to deal specifically with the problem of fine-grained classes ?\n\nCheers"
  }
}