{
  "id": 401924,
  "title": "How is the competition data normalized?",
  "url": "/competitions/asl-signs/discussion/401924",
  "author_name": "",
  "post_date": "2023-04-15T18:26:46.811179100Z",
  "votes": null,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Hi Everyone,</p>\n<p>I'm working on trying to improve my models with augmentation and normalization today and I'm curious, does anyone know how the data in this competition was normalized? The paragraph below is from the data description page of the competition, but it doesn't say <em>how</em> the data was normalized. I've seen normalization discussed quite a lot in discussions as a way to improve scores, but I'm not completely sure what kind of normalization we're starting with from the raw competition data. <a href=\"https://www.kaggle.com/hengck\" target=\"_blank\">@hengck</a> you have mentioned normalization quite a lot in your many discussion posts, perhaps you know how the data is already normalized?</p>\n<p>From the competition's data page:<br>\n\"[x/y/z] - The normalized spatial coordinates of the landmark. These are the only columns that will be provided to your submitted model for inference. The MediaPipe model is not fully trained to predict depth so you may wish to ignore the z values.\"</p>",
  "messages": [
    {
      "id": "2223028",
      "postDate": "04/15/2023 18:26:46",
      "content": "<p>Hi Everyone,</p>\n<p>I'm working on trying to improve my models with augmentation and normalization today and I'm curious, does anyone know how the data in this competition was normalized? The paragraph below is from the data description page of the competition, but it doesn't say <em>how</em> the data was normalized. I've seen normalization discussed quite a lot in discussions as a way to improve scores, but I'm not completely sure what kind of normalization we're starting with from the raw competition data. <a href=\"https://www.kaggle.com/hengck\" target=\"_blank\">@hengck</a> you have mentioned normalization quite a lot in your many discussion posts, perhaps you know how the data is already normalized?</p>\n<p>From the competition's data page:<br>\n\"[x/y/z] - The normalized spatial coordinates of the landmark. These are the only columns that will be provided to your submitted model for inference. The MediaPipe model is not fully trained to predict depth so you may wish to ignore the z values.\"</p>",
      "rawMarkdown": "Hi Everyone,\n\nI'm working on trying to improve my models with augmentation and normalization today and I'm curious, does anyone know how the data in this competition was normalized? The paragraph below is from the data description page of the competition, but it doesn't say *how* the data was normalized. I've seen normalization discussed quite a lot in discussions as a way to improve scores, but I'm not completely sure what kind of normalization we're starting with from the raw competition data. @hengck you have mentioned normalization quite a lot in your many discussion posts, perhaps you know how the data is already normalized?\n\nFrom the competition's data page:\n\"[x/y/z] - The normalized spatial coordinates of the landmark. These are the only columns that will be provided to your submitted model for inference. The MediaPipe model is not fully trained to predict depth so you may wish to ignore the z values.\"",
      "votes": null
    },
    {
      "id": "2224781",
      "postDate": "04/17/2023 16:46:09",
      "content": "<p>We just used MediaPipe's normalization.</p>",
      "rawMarkdown": "We just used MediaPipe's normalization.",
      "votes": null
    },
    {
      "id": "2225157",
      "postDate": "04/18/2023 00:18:05",
      "content": "<p>Thanks <a href=\"https://www.kaggle.com/sohier\" target=\"_blank\">@sohier</a> </p>",
      "rawMarkdown": "Thanks @sohier",
      "votes": null
    },
    {
      "id": "2225246",
      "postDate": "04/18/2023 02:43:07",
      "content": "<p>👀👀👀👀👀👀👀👀👀👀👀👀👀👀👀👀👀</p>",
      "rawMarkdown": "👀👀👀👀👀👀👀👀👀👀👀👀👀👀👀👀👀",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2224781,
      "author_name": "sohier",
      "author_url": "",
      "post_date": "04/17/2023 16:46:09",
      "content": "<p>We just used MediaPipe's normalization.</p>",
      "votes": null,
      "replies": [
        {
          "id": 2225157,
          "author_name": "chemdatafarmer",
          "author_url": "",
          "post_date": "04/18/2023 00:18:05",
          "content": "<p>Thanks <a href=\"https://www.kaggle.com/sohier\" target=\"_blank\">@sohier</a> </p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 2225246,
      "author_name": "lemonnice",
      "author_url": "",
      "post_date": "04/18/2023 02:43:07",
      "content": "<p>👀👀👀👀👀👀👀👀👀👀👀👀👀👀👀👀👀</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2223028": "Hi Everyone,\n\nI'm working on trying to improve my models with augmentation and normalization today and I'm curious, does anyone know how the data in this competition was normalized? The paragraph below is from the data description page of the competition, but it doesn't say *how* the data was normalized. I've seen normalization discussed quite a lot in discussions as a way to improve scores, but I'm not completely sure what kind of normalization we're starting with from the raw competition data. @hengck you have mentioned normalization quite a lot in your many discussion posts, perhaps you know how the data is already normalized?\n\nFrom the competition's data page:\n\"[x/y/z] - The normalized spatial coordinates of the landmark. These are the only columns that will be provided to your submitted model for inference. The MediaPipe model is not fully trained to predict depth so you may wish to ignore the z values.\"",
    "2224781": "We just used MediaPipe's normalization.",
    "2225157": "Thanks @sohier",
    "2225246": "👀👀👀👀👀👀👀👀👀👀👀👀👀👀👀👀👀"
  },
  "source": "meta"
}