{
  "id": 134691,
  "title": "Welcome to Semi-Supervised Fine-Grained Recognition Challenge!",
  "url": "/competitions/semi-inat-2020/discussion/134691",
  "author_name": "",
  "post_date": "2020-03-09T19:42:20.080340400Z",
  "votes": 3,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Hi everyone,</p>\n\n<p>We are happy to announce the first semi-supervised fine-grained recognition (iNaturalist Aves) challenge. The challenge is part of the Seventh Workshop on Fine-Grained Visual Categorization (<a href=\"https://sites.google.com/view/fgvc7\">FGVC7</a>) at CVPR 2020 in Seattle.</p>\n\n<p>Different from other semi-supervised benchmarks (e.g. 10% ImageNet or cifar/svhn), we focus on (1) long-tailed distribution of classes, (2) the unlabeled data from both target and non-target classes, and (3) fine-grained similarity between classes.</p>\n\n<p>Looking forward to the interesting and creative solutions that you all come up with. Feel free to ask questions here, or on the <a href=\"https://github.com/cvl-umass/semi-inat-2020\">github page</a>!</p>",
  "messages": [
    {
      "id": "767550",
      "postDate": "03/09/2020 19:42:20",
      "content": "<p>Hi everyone,</p>\n\n<p>We are happy to announce the first semi-supervised fine-grained recognition (iNaturalist Aves) challenge. The challenge is part of the Seventh Workshop on Fine-Grained Visual Categorization (<a href=\"https://sites.google.com/view/fgvc7\">FGVC7</a>) at CVPR 2020 in Seattle.</p>\n\n<p>Different from other semi-supervised benchmarks (e.g. 10% ImageNet or cifar/svhn), we focus on (1) long-tailed distribution of classes, (2) the unlabeled data from both target and non-target classes, and (3) fine-grained similarity between classes.</p>\n\n<p>Looking forward to the interesting and creative solutions that you all come up with. Feel free to ask questions here, or on the <a href=\"https://github.com/cvl-umass/semi-inat-2020\">github page</a>!</p>",
      "rawMarkdown": "Hi everyone,\n\nWe are happy to announce the first semi-supervised fine-grained recognition (iNaturalist Aves) challenge. The challenge is part of the Seventh Workshop on Fine-Grained Visual Categorization ([FGVC7](https://sites.google.com/view/fgvc7)) at CVPR 2020 in Seattle.\n\nDifferent from other semi-supervised benchmarks (e.g. 10% ImageNet or cifar/svhn), we focus on (1) long-tailed distribution of classes, (2) the unlabeled data from both target and non-target classes, and (3) fine-grained similarity between classes.\n\nLooking forward to the interesting and creative solutions that you all come up with. Feel free to ask questions here, or on the [github page](https://github.com/cvl-umass/semi-inat-2020)!",
      "votes": null
    },
    {
      "id": "770120",
      "postDate": "03/12/2020 15:34:14",
      "content": "<p>Hi, thanks for the interesting problem.</p>\n\n<p>I'd like to ask if we are allowed to use pretrained object detection networks like YOLO?</p>",
      "rawMarkdown": "Hi, thanks for the interesting problem.\n\nI'd like to ask if we are allowed to use pretrained object detection networks like YOLO?",
      "votes": null
    },
    {
      "id": "770543",
      "postDate": "03/13/2020 04:29:21",
      "content": "<p>Hi, other pre-trained networks like YOLO are not allowed, because it was trained on dataset other than ImageNet. The only thing allowed is the ImageNet pre-trained model. It is because we would like people to focus on semi-supervised methods rather than using more training data. Thanks!</p>",
      "rawMarkdown": "Hi, other pre-trained networks like YOLO are not allowed, because it was trained on dataset other than ImageNet. The only thing allowed is the ImageNet pre-trained model. It is because we would like people to focus on semi-supervised methods rather than using more training data. Thanks!",
      "votes": null
    },
    {
      "id": "770930",
      "postDate": "03/13/2020 15:24:22",
      "content": "<p>I see, thanks!</p>",
      "rawMarkdown": "I see, thanks!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 770120,
      "author_name": "ananschuett",
      "author_url": "",
      "post_date": "03/12/2020 15:34:14",
      "content": "<p>Hi, thanks for the interesting problem.</p>\n\n<p>I'd like to ask if we are allowed to use pretrained object detection networks like YOLO?</p>",
      "votes": null,
      "replies": [
        {
          "id": 770543,
          "author_name": "jcfredsu",
          "author_url": "",
          "post_date": "03/13/2020 04:29:21",
          "content": "<p>Hi, other pre-trained networks like YOLO are not allowed, because it was trained on dataset other than ImageNet. The only thing allowed is the ImageNet pre-trained model. It is because we would like people to focus on semi-supervised methods rather than using more training data. Thanks!</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 770930,
          "author_name": "ananschuett",
          "author_url": "",
          "post_date": "03/13/2020 15:24:22",
          "content": "<p>I see, thanks!</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "767550": "Hi everyone,\n\nWe are happy to announce the first semi-supervised fine-grained recognition (iNaturalist Aves) challenge. The challenge is part of the Seventh Workshop on Fine-Grained Visual Categorization ([FGVC7](https://sites.google.com/view/fgvc7)) at CVPR 2020 in Seattle.\n\nDifferent from other semi-supervised benchmarks (e.g. 10% ImageNet or cifar/svhn), we focus on (1) long-tailed distribution of classes, (2) the unlabeled data from both target and non-target classes, and (3) fine-grained similarity between classes.\n\nLooking forward to the interesting and creative solutions that you all come up with. Feel free to ask questions here, or on the [github page](https://github.com/cvl-umass/semi-inat-2020)!",
    "770120": "Hi, thanks for the interesting problem.\n\nI'd like to ask if we are allowed to use pretrained object detection networks like YOLO?",
    "770543": "Hi, other pre-trained networks like YOLO are not allowed, because it was trained on dataset other than ImageNet. The only thing allowed is the ImageNet pre-trained model. It is because we would like people to focus on semi-supervised methods rather than using more training data. Thanks!",
    "770930": "I see, thanks!"
  },
  "source": "meta"
}