{
  "id": 252237,
  "title": "EfficientNet for Classification and as EfficientDet Backbone (Multi Head)",
  "url": "/competitions/siim-covid19-detection/discussion/252237",
  "author_name": "",
  "post_date": "2021-07-11T09:01:05.433582100Z",
  "votes": 2,
  "comment_count": 5,
  "views": 0,
  "content": "<p>Has anyone managed to simultaneously train a classifier EfficientNet and detection EfficientDet with the same backbone? Does anyone have an idea to do this? (The theory is obvious but im not sure how to go about training it in practice) I guess Tensorflow might be superior here but I really don't know. I've been doing some experimenting with rosswrightman timm and effdet but I've only been able to make the nets work independently.</p>",
  "messages": [
    {
      "id": "1383803",
      "postDate": "07/11/2021 09:01:05",
      "content": "<p>Has anyone managed to simultaneously train a classifier EfficientNet and detection EfficientDet with the same backbone? Does anyone have an idea to do this? (The theory is obvious but im not sure how to go about training it in practice) I guess Tensorflow might be superior here but I really don't know. I've been doing some experimenting with rosswrightman timm and effdet but I've only been able to make the nets work independently.</p>",
      "rawMarkdown": "Has anyone managed to simultaneously train a classifier EfficientNet and detection EfficientDet with the same backbone? Does anyone have an idea to do this? (The theory is obvious but im not sure how to go about training it in practice) I guess Tensorflow might be superior here but I really don't know. I've been doing some experimenting with rosswrightman timm and effdet but I've only been able to make the nets work independently.",
      "votes": null
    },
    {
      "id": "1387610",
      "postDate": "07/14/2021 10:00:17",
      "content": "<p>Hello, <a href=\"https://www.kaggle.com/simon111\" target=\"_blank\">@simon111</a> , I don't know if I understand you correctly, you mean to use multi-tasking learning to simultaneously complete classification and detection task in a network, right?</p>",
      "rawMarkdown": "Hello, @simon111 , I don't know if I understand you correctly, you mean to use multi-tasking learning to simultaneously complete classification and detection task in a network, right?",
      "votes": null
    },
    {
      "id": "1387623",
      "postDate": "07/14/2021 10:14:41",
      "content": "<p>Yes I mean basically training EfficientDet and EfficientNet at the same time with one shared backbone</p>",
      "rawMarkdown": "Yes I mean basically training EfficientDet and EfficientNet at the same time with one shared backbone",
      "votes": null
    },
    {
      "id": "1387821",
      "postDate": "07/14/2021 12:53:21",
      "content": "<p>Good idea, it worth trying!</p>",
      "rawMarkdown": "Good idea, it worth trying!",
      "votes": null
    },
    {
      "id": "1387938",
      "postDate": "07/14/2021 14:26:45",
      "content": "<p>Do you have any idea how it could be done?</p>",
      "rawMarkdown": "Do you have any idea how it could be done?",
      "votes": null
    },
    {
      "id": "1388469",
      "postDate": "07/15/2021 01:57:48",
      "content": "<p>I am not familar with EfficientDet, so it will take some time for me to learn EfficientDet, and then evaluating the feasibility of training EfficientDet and EfficientNet at the same time with one shared backbone.</p>",
      "rawMarkdown": "I am not familar with EfficientDet, so it will take some time for me to learn EfficientDet, and then evaluating the feasibility of training EfficientDet and EfficientNet at the same time with one shared backbone.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1387610,
      "author_name": "shangweichen",
      "author_url": "",
      "post_date": "07/14/2021 10:00:17",
      "content": "<p>Hello, <a href=\"https://www.kaggle.com/simon111\" target=\"_blank\">@simon111</a> , I don't know if I understand you correctly, you mean to use multi-tasking learning to simultaneously complete classification and detection task in a network, right?</p>",
      "votes": null,
      "replies": [
        {
          "id": 1387623,
          "author_name": "simon111",
          "author_url": "",
          "post_date": "07/14/2021 10:14:41",
          "content": "<p>Yes I mean basically training EfficientDet and EfficientNet at the same time with one shared backbone</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1387821,
          "author_name": "shangweichen",
          "author_url": "",
          "post_date": "07/14/2021 12:53:21",
          "content": "<p>Good idea, it worth trying!</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1387938,
          "author_name": "simon111",
          "author_url": "",
          "post_date": "07/14/2021 14:26:45",
          "content": "<p>Do you have any idea how it could be done?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1388469,
          "author_name": "shangweichen",
          "author_url": "",
          "post_date": "07/15/2021 01:57:48",
          "content": "<p>I am not familar with EfficientDet, so it will take some time for me to learn EfficientDet, and then evaluating the feasibility of training EfficientDet and EfficientNet at the same time with one shared backbone.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1383803": "Has anyone managed to simultaneously train a classifier EfficientNet and detection EfficientDet with the same backbone? Does anyone have an idea to do this? (The theory is obvious but im not sure how to go about training it in practice) I guess Tensorflow might be superior here but I really don't know. I've been doing some experimenting with rosswrightman timm and effdet but I've only been able to make the nets work independently.",
    "1387610": "Hello, @simon111 , I don't know if I understand you correctly, you mean to use multi-tasking learning to simultaneously complete classification and detection task in a network, right?",
    "1387623": "Yes I mean basically training EfficientDet and EfficientNet at the same time with one shared backbone",
    "1387821": "Good idea, it worth trying!",
    "1387938": "Do you have any idea how it could be done?",
    "1388469": "I am not familar with EfficientDet, so it will take some time for me to learn EfficientDet, and then evaluating the feasibility of training EfficientDet and EfficientNet at the same time with one shared backbone."
  },
  "source": "meta"
}