{
  "id": 313996,
  "title": "Are there any potential NN structure other than effnet+arcface?",
  "url": "/competitions/happy-whale-and-dolphin/discussion/313996",
  "author_name": "",
  "post_date": "2022-03-20T10:17:46.589199500Z",
  "votes": 4,
  "comment_count": 4,
  "views": 0,
  "content": "<p>I see almost all the good baselines are using effnet + arcface neural network structure. I am wondering if other structures could be also helpful. I appreciate that you could give some thoughts.</p>",
  "messages": [
    {
      "id": "1729644",
      "postDate": "03/20/2022 10:17:46",
      "content": "<p>I see almost all the good baselines are using effnet + arcface neural network structure. I am wondering if other structures could be also helpful. I appreciate that you could give some thoughts.</p>",
      "rawMarkdown": "I see almost all the good baselines are using effnet + arcface neural network structure. I am wondering if other structures could be also helpful. I appreciate that you could give some thoughts.",
      "votes": null
    },
    {
      "id": "1729669",
      "postDate": "03/20/2022 10:38:07",
      "content": "<p>I suggest you reading this survey: <a href=\"https://arxiv.org/abs/1804.06655\" target=\"_blank\">https://arxiv.org/abs/1804.06655</a> There are a lot of attitude we can use. </p>",
      "rawMarkdown": "I suggest you reading this survey: https://arxiv.org/abs/1804.06655 There are a lot of attitude we can use.",
      "votes": null
    },
    {
      "id": "1729968",
      "postDate": "03/20/2022 18:31:18",
      "content": "<p>Thank you! I will take a deeper look into that and share what I found.</p>",
      "rawMarkdown": "Thank you! I will take a deeper look into that and share what I found.",
      "votes": null
    },
    {
      "id": "1730427",
      "postDate": "03/21/2022 08:36:46",
      "content": "<p>I use ConvNeXt + ArcFace instead of EfficientNet and got a good boost (using PyTorch): <a href=\"https://www.kaggle.com/code/clemchris/pytorch-backfin-convnext-arcface/notebook\" target=\"_blank\">https://www.kaggle.com/code/clemchris/pytorch-backfin-convnext-arcface/notebook</a>.</p>",
      "rawMarkdown": "I use ConvNeXt + ArcFace instead of EfficientNet and got a good boost (using PyTorch): https://www.kaggle.com/code/clemchris/pytorch-backfin-convnext-arcface/notebook.",
      "votes": null
    },
    {
      "id": "1730497",
      "postDate": "03/21/2022 10:42:31",
      "content": "<p>Nice notebook! I will try and see if I can do that on TensorFlow.</p>",
      "rawMarkdown": "Nice notebook! I will try and see if I can do that on TensorFlow.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1729669,
      "author_name": "remekkinas",
      "author_url": "",
      "post_date": "03/20/2022 10:38:07",
      "content": "<p>I suggest you reading this survey: <a href=\"https://arxiv.org/abs/1804.06655\" target=\"_blank\">https://arxiv.org/abs/1804.06655</a> There are a lot of attitude we can use. </p>",
      "votes": null,
      "replies": [
        {
          "id": 1729968,
          "author_name": "runjiali",
          "author_url": "",
          "post_date": "03/20/2022 18:31:18",
          "content": "<p>Thank you! I will take a deeper look into that and share what I found.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1730427,
      "author_name": "clemchris",
      "author_url": "",
      "post_date": "03/21/2022 08:36:46",
      "content": "<p>I use ConvNeXt + ArcFace instead of EfficientNet and got a good boost (using PyTorch): <a href=\"https://www.kaggle.com/code/clemchris/pytorch-backfin-convnext-arcface/notebook\" target=\"_blank\">https://www.kaggle.com/code/clemchris/pytorch-backfin-convnext-arcface/notebook</a>.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1730497,
          "author_name": "runjiali",
          "author_url": "",
          "post_date": "03/21/2022 10:42:31",
          "content": "<p>Nice notebook! I will try and see if I can do that on TensorFlow.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1729644": "I see almost all the good baselines are using effnet + arcface neural network structure. I am wondering if other structures could be also helpful. I appreciate that you could give some thoughts.",
    "1729669": "I suggest you reading this survey: https://arxiv.org/abs/1804.06655 There are a lot of attitude we can use.",
    "1729968": "Thank you! I will take a deeper look into that and share what I found.",
    "1730427": "I use ConvNeXt + ArcFace instead of EfficientNet and got a good boost (using PyTorch): https://www.kaggle.com/code/clemchris/pytorch-backfin-convnext-arcface/notebook.",
    "1730497": "Nice notebook! I will try and see if I can do that on TensorFlow."
  },
  "source": "meta"
}