{
  "id": 285356,
  "title": "Private Dataset with model",
  "url": "/competitions/sartorius-cell-instance-segmentation/discussion/285356",
  "author_name": "",
  "post_date": "2021-11-04T11:11:11.445879Z",
  "votes": 2,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Hello!</p>\n<p>\"Internet access disabled<br>\nFreely &amp; publicly available external data is allowed, including pre-trained models\"</p>\n<p>Can I: </p>\n<ol>\n<li>train model</li>\n<li>store model it in my private dataset </li>\n<li>for submition use model from dataset?</li>\n</ol>\n<p>Or I should store model in public dataset avaible for all members?</p>",
  "messages": [
    {
      "id": "1570768",
      "postDate": "11/04/2021 11:11:11",
      "content": "<p>Hello!</p>\n<p>\"Internet access disabled<br>\nFreely &amp; publicly available external data is allowed, including pre-trained models\"</p>\n<p>Can I: </p>\n<ol>\n<li>train model</li>\n<li>store model it in my private dataset </li>\n<li>for submition use model from dataset?</li>\n</ol>\n<p>Or I should store model in public dataset avaible for all members?</p>",
      "rawMarkdown": "Hello!\n\n\"Internet access disabled\nFreely & publicly available external data is allowed, including pre-trained models\"\n\nCan I: \n1. train model\n2. store model it in my private dataset \n3. for submition use model from dataset?\n\nOr I should store model in public dataset avaible for all members?",
      "votes": null
    },
    {
      "id": "1570827",
      "postDate": "11/04/2021 12:06:38",
      "content": "<p>What you are describing is the most common way to compete.</p>\n<p>Make a model<br>\nStore it in a private dataset<br>\nSubmit inference based on that model.</p>\n<p>As long as your model uses only the competition dataset or other publicly available data, you do not need to make your model available to others. Also, any pre-trained weights or software you start with must be publicly available as described in the rules.</p>\n<p>Earlier contests often required you to \"publicly declare\" outside resources. That no longer seems to be required. Just that they are available. You do not have to point them out to everybody.</p>\n<p>-Rich</p>",
      "rawMarkdown": "What you are describing is the most common way to compete.\n\nMake a model\nStore it in a private dataset\nSubmit inference based on that model.\n\nAs long as your model uses only the competition dataset or other publicly available data, you do not need to make your model available to others. Also, any pre-trained weights or software you start with must be publicly available as described in the rules.\n\nEarlier contests often required you to \"publicly declare\" outside resources. That no longer seems to be required. Just that they are available. You do not have to point them out to everybody.\n\n-Rich",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1570827,
      "author_name": "richardepstein",
      "author_url": "",
      "post_date": "11/04/2021 12:06:38",
      "content": "<p>What you are describing is the most common way to compete.</p>\n<p>Make a model<br>\nStore it in a private dataset<br>\nSubmit inference based on that model.</p>\n<p>As long as your model uses only the competition dataset or other publicly available data, you do not need to make your model available to others. Also, any pre-trained weights or software you start with must be publicly available as described in the rules.</p>\n<p>Earlier contests often required you to \"publicly declare\" outside resources. That no longer seems to be required. Just that they are available. You do not have to point them out to everybody.</p>\n<p>-Rich</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1570768": "Hello!\n\n\"Internet access disabled\nFreely & publicly available external data is allowed, including pre-trained models\"\n\nCan I: \n1. train model\n2. store model it in my private dataset \n3. for submition use model from dataset?\n\nOr I should store model in public dataset avaible for all members?",
    "1570827": "What you are describing is the most common way to compete.\n\nMake a model\nStore it in a private dataset\nSubmit inference based on that model.\n\nAs long as your model uses only the competition dataset or other publicly available data, you do not need to make your model available to others. Also, any pre-trained weights or software you start with must be publicly available as described in the rules.\n\nEarlier contests often required you to \"publicly declare\" outside resources. That no longer seems to be required. Just that they are available. You do not have to point them out to everybody.\n\n-Rich"
  },
  "source": "meta"
}