{
  "id": 452718,
  "title": "Is pretrained model allowed?",
  "url": "/competitions/aocr2024/discussion/452718",
  "author_name": "",
  "post_date": "2023-11-03T08:49:28.778015800Z",
  "votes": null,
  "comment_count": 6,
  "views": 0,
  "content": "<p>I've read through the rules but didn't see any mentions about pretrained model policy. Am I allowed for use pretrained model, for example ResNet-50 on ImageNet-1k, for initialization of the network?</p>\n<p>Thank you very much.</p>",
  "messages": [
    {
      "id": "2510785",
      "postDate": "11/03/2023 08:49:28",
      "content": "<p>I've read through the rules but didn't see any mentions about pretrained model policy. Am I allowed for use pretrained model, for example ResNet-50 on ImageNet-1k, for initialization of the network?</p>\n<p>Thank you very much.</p>",
      "rawMarkdown": "I've read through the rules but didn't see any mentions about pretrained model policy. Am I allowed for use pretrained model, for example ResNet-50 on ImageNet-1k, for initialization of the network?\n\nThank you very much.",
      "votes": null
    },
    {
      "id": "2510809",
      "postDate": "11/03/2023 09:07:30",
      "content": "<p>Yes, you are allowed to use pre-trained models, such as ResNet-50 initialized on ImageNet-1k, to initialize your network. However, it's essential to ensure that any pre-trained model you use is generally available to all participants. Please adhere to this guideline and all other aspects of the competition rules.</p>",
      "rawMarkdown": "Yes, you are allowed to use pre-trained models, such as ResNet-50 initialized on ImageNet-1k, to initialize your network. However, it's essential to ensure that any pre-trained model you use is generally available to all participants. Please adhere to this guideline and all other aspects of the competition rules.",
      "votes": null
    },
    {
      "id": "2510812",
      "postDate": "11/03/2023 09:09:51",
      "content": "<p>Thank you very much for your answer.</p>",
      "rawMarkdown": "Thank you very much for your answer.",
      "votes": null
    },
    {
      "id": "2518044",
      "postDate": "11/09/2023 02:06:06",
      "content": "<p>If using general purpose pretrained model, such as \"ImageNet\", that is definitely allowed. <br>\nHowever, if using specific purpose model related to \"acute abdomen or appendicitis\" as pretrained weight, then the cases of training the model should be open to public. </p>",
      "rawMarkdown": "If using general purpose pretrained model, such as \"ImageNet\", that is definitely allowed. \nHowever, if using specific purpose model related to \"acute abdomen or appendicitis\" as pretrained weight, then the cases of training the model should be open to public.",
      "votes": null
    },
    {
      "id": "2518048",
      "postDate": "11/09/2023 02:19:31",
      "content": "<p>We are using a public dataset of CT scans of various body parts (head, chest, abdomen) for pretraining. This helps the model having an intuition of what a CT slices look like because models pretrained on ImageNet aren't that familiar with CT scans.</p>\n<p>Just to be clear, do we have to disclose which datasets we are using \"during\" the competition period? Given that these datasets and pretrained models are publicly available, anyone can download without registration. These datasets aren't related to \"acute abdomen or appendicitis\". </p>",
      "rawMarkdown": "We are using a public dataset of CT scans of various body parts (head, chest, abdomen) for pretraining. This helps the model having an intuition of what a CT slices look like because models pretrained on ImageNet aren't that familiar with CT scans.\n\nJust to be clear, do we have to disclose which datasets we are using \"during\" the competition period? Given that these datasets and pretrained models are publicly available, anyone can download without registration. These datasets aren't related to \"acute abdomen or appendicitis\".",
      "votes": null
    },
    {
      "id": "2518058",
      "postDate": "11/09/2023 02:40:06",
      "content": "<ol>\n<li>If this model is not directly related to disease, such as anatomy model, it is allowed. : )<br>\n2.If using extra data related to acute abdomen disease, competitor should share the data or share the link of original source. <br>\nCompetitor does not need to disclose this information \"during\" the competition period, but after the competition ends, if you finish in the top five, you need to share above information.</li>\n</ol>",
      "rawMarkdown": "1. If this model is not directly related to disease, such as anatomy model, it is allowed. : )\n2.If using extra data related to acute abdomen disease, competitor should share the data or share the link of original source. \nCompetitor does not need to disclose this information \"during\" the competition period, but after the competition ends, if you finish in the top five, you need to share above information.",
      "votes": null
    },
    {
      "id": "2518059",
      "postDate": "11/09/2023 02:42:02",
      "content": "<p>Thank you for your clarification. </p>",
      "rawMarkdown": "Thank you for your clarification.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2510809,
      "author_name": "wenjeng",
      "author_url": "",
      "post_date": "11/03/2023 09:07:30",
      "content": "<p>Yes, you are allowed to use pre-trained models, such as ResNet-50 initialized on ImageNet-1k, to initialize your network. However, it's essential to ensure that any pre-trained model you use is generally available to all participants. Please adhere to this guideline and all other aspects of the competition rules.</p>",
      "votes": null,
      "replies": [
        {
          "id": 2510812,
          "author_name": "thangngoc89",
          "author_url": "",
          "post_date": "11/03/2023 09:09:51",
          "content": "<p>Thank you very much for your answer.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 2518044,
      "author_name": "goman178",
      "author_url": "",
      "post_date": "11/09/2023 02:06:06",
      "content": "<p>If using general purpose pretrained model, such as \"ImageNet\", that is definitely allowed. <br>\nHowever, if using specific purpose model related to \"acute abdomen or appendicitis\" as pretrained weight, then the cases of training the model should be open to public. </p>",
      "votes": null,
      "replies": [
        {
          "id": 2518048,
          "author_name": "thangngoc89",
          "author_url": "",
          "post_date": "11/09/2023 02:19:31",
          "content": "<p>We are using a public dataset of CT scans of various body parts (head, chest, abdomen) for pretraining. This helps the model having an intuition of what a CT slices look like because models pretrained on ImageNet aren't that familiar with CT scans.</p>\n<p>Just to be clear, do we have to disclose which datasets we are using \"during\" the competition period? Given that these datasets and pretrained models are publicly available, anyone can download without registration. These datasets aren't related to \"acute abdomen or appendicitis\". </p>",
          "votes": null,
          "replies": [
            {
              "id": 2518058,
              "author_name": "goman178",
              "author_url": "",
              "post_date": "11/09/2023 02:40:06",
              "content": "<ol>\n<li>If this model is not directly related to disease, such as anatomy model, it is allowed. : )<br>\n2.If using extra data related to acute abdomen disease, competitor should share the data or share the link of original source. <br>\nCompetitor does not need to disclose this information \"during\" the competition period, but after the competition ends, if you finish in the top five, you need to share above information.</li>\n</ol>",
              "votes": null,
              "replies": [
                {
                  "id": 2518059,
                  "author_name": "thangngoc89",
                  "author_url": "",
                  "post_date": "11/09/2023 02:42:02",
                  "content": "<p>Thank you for your clarification. </p>",
                  "votes": null,
                  "replies": []
                }
              ]
            }
          ]
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2510785": "I've read through the rules but didn't see any mentions about pretrained model policy. Am I allowed for use pretrained model, for example ResNet-50 on ImageNet-1k, for initialization of the network?\n\nThank you very much.",
    "2510809": "Yes, you are allowed to use pre-trained models, such as ResNet-50 initialized on ImageNet-1k, to initialize your network. However, it's essential to ensure that any pre-trained model you use is generally available to all participants. Please adhere to this guideline and all other aspects of the competition rules.",
    "2510812": "Thank you very much for your answer.",
    "2518044": "If using general purpose pretrained model, such as \"ImageNet\", that is definitely allowed. \nHowever, if using specific purpose model related to \"acute abdomen or appendicitis\" as pretrained weight, then the cases of training the model should be open to public.",
    "2518048": "We are using a public dataset of CT scans of various body parts (head, chest, abdomen) for pretraining. This helps the model having an intuition of what a CT slices look like because models pretrained on ImageNet aren't that familiar with CT scans.\n\nJust to be clear, do we have to disclose which datasets we are using \"during\" the competition period? Given that these datasets and pretrained models are publicly available, anyone can download without registration. These datasets aren't related to \"acute abdomen or appendicitis\".",
    "2518058": "1. If this model is not directly related to disease, such as anatomy model, it is allowed. : )\n2.If using extra data related to acute abdomen disease, competitor should share the data or share the link of original source. \nCompetitor does not need to disclose this information \"during\" the competition period, but after the competition ends, if you finish in the top five, you need to share above information.",
    "2518059": "Thank you for your clarification."
  },
  "source": "meta"
}