{
  "id": 511734,
  "title": "Usage/creation of train data.",
  "url": "/competitions/rsna-2024-lumbar-spine-degenerative-classification/discussion/511734",
  "author_name": "",
  "post_date": "2024-06-11T20:50:31.114336600Z",
  "votes": null,
  "comment_count": 3,
  "views": 0,
  "content": "<p>I'm presuming this is not possible, but are we confined to using the fixed train data alone in developing our model ? Or are various data augmentation techniques possible (whether on / not on the fly come test time ?</p>",
  "messages": [
    {
      "id": "2867469",
      "postDate": "06/11/2024 20:50:31",
      "content": "<p>I'm presuming this is not possible, but are we confined to using the fixed train data alone in developing our model ? Or are various data augmentation techniques possible (whether on / not on the fly come test time ?</p>",
      "rawMarkdown": "I'm presuming this is not possible, but are we confined to using the fixed train data alone in developing our model ? Or are various data augmentation techniques possible (whether on / not on the fly come test time ?",
      "votes": null
    },
    {
      "id": "2867632",
      "postDate": "06/12/2024 01:42:07",
      "content": "<p>Based on rules, public external dataset is allowed.</p>",
      "rawMarkdown": "Based on rules, public external dataset is allowed.",
      "votes": null
    },
    {
      "id": "2874180",
      "postDate": "06/16/2024 05:35:41",
      "content": "<p>I'm presuming this is not possible, but are we confined to using the fixed train data alone in developing our model ? </p>\n<p>Make/upload your own dataset (or any dataset) and set it to public. This way, I think, you can use them.</p>",
      "rawMarkdown": "I'm presuming this is not possible, but are we confined to using the fixed train data alone in developing our model ? \n\nMake/upload your own dataset (or any dataset) and set it to public. This way, I think, you can use them.",
      "votes": null
    },
    {
      "id": "2874188",
      "postDate": "06/16/2024 05:40:51",
      "content": "<p>Or are various data augmentation techniques possible (whether on / not on the fly come test time?</p>\n<p>Yes, you can have data augmentation layers and functions in the train time. You can also on or off them, in the test time or, while submitting the notebook for evaluation. I suggest you to on them while you are training the the model but off them when the model is making predictions for the final submission</p>",
      "rawMarkdown": "Or are various data augmentation techniques possible (whether on / not on the fly come test time?\n\nYes, you can have data augmentation layers and functions in the train time. You can also on or off them, in the test time or, while submitting the notebook for evaluation. I suggest you to on them while you are training the the model but off them when the model is making predictions for the final submission",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2867632,
      "author_name": "shipuxu",
      "author_url": "",
      "post_date": "06/12/2024 01:42:07",
      "content": "<p>Based on rules, public external dataset is allowed.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 2874180,
      "author_name": "devsya",
      "author_url": "",
      "post_date": "06/16/2024 05:35:41",
      "content": "<p>I'm presuming this is not possible, but are we confined to using the fixed train data alone in developing our model ? </p>\n<p>Make/upload your own dataset (or any dataset) and set it to public. This way, I think, you can use them.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 2874188,
      "author_name": "devsya",
      "author_url": "",
      "post_date": "06/16/2024 05:40:51",
      "content": "<p>Or are various data augmentation techniques possible (whether on / not on the fly come test time?</p>\n<p>Yes, you can have data augmentation layers and functions in the train time. You can also on or off them, in the test time or, while submitting the notebook for evaluation. I suggest you to on them while you are training the the model but off them when the model is making predictions for the final submission</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2867469": "I'm presuming this is not possible, but are we confined to using the fixed train data alone in developing our model ? Or are various data augmentation techniques possible (whether on / not on the fly come test time ?",
    "2867632": "Based on rules, public external dataset is allowed.",
    "2874180": "I'm presuming this is not possible, but are we confined to using the fixed train data alone in developing our model ? \n\nMake/upload your own dataset (or any dataset) and set it to public. This way, I think, you can use them.",
    "2874188": "Or are various data augmentation techniques possible (whether on / not on the fly come test time?\n\nYes, you can have data augmentation layers and functions in the train time. You can also on or off them, in the test time or, while submitting the notebook for evaluation. I suggest you to on them while you are training the the model but off them when the model is making predictions for the final submission"
  },
  "source": "meta"
}