{
  "id": 392733,
  "title": "clarifications about the moidel ?",
  "url": "/competitions/asl-signs/discussion/392733",
  "author_name": "",
  "post_date": "2023-03-06T16:38:10.661806700Z",
  "votes": 1,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hi everyone  </p>\n<p>I want to know if we must build the model from scratch ,also if you have some propostion of what architecture to use ?<br>\nand for the input shape is it equal to [543,3] or I am wrong ?</p>\n<p>thanks for your help</p>",
  "messages": [
    {
      "id": "2171282",
      "postDate": "03/06/2023 16:38:10",
      "content": "<p>Hi everyone  </p>\n<p>I want to know if we must build the model from scratch ,also if you have some propostion of what architecture to use ?<br>\nand for the input shape is it equal to [543,3] or I am wrong ?</p>\n<p>thanks for your help</p>",
      "rawMarkdown": "Hi everyone  \n\nI want to know if we must build the model from scratch ,also if you have some propostion of what architecture to use ?\nand for the input shape is it equal to [543,3] or I am wrong ?\n\nthanks for your help",
      "votes": null
    },
    {
      "id": "2171532",
      "postDate": "03/06/2023 21:31:19",
      "content": "<p>There's nothing that says you have to build a model a from scratch, but that's what all the top-scoring notebooks have done so far. <br>\n There are various different architectures you might want to use.  Take a look at some of the public notebooks that have been shared and you'll see examples of (at least), DNNs, LSTMs &amp; Transformer networks.</p>\n<p>The input shape is <code>(None, 543, 3)</code> (where the initial <code>None</code> is a placeholder for a variable number of frames, the 543 is the landmarks and the 3 is the x/y/z components).</p>",
      "rawMarkdown": "There's nothing that says you have to build a model a from scratch, but that's what all the top-scoring notebooks have done so far. \n There are various different architectures you might want to use.  Take a look at some of the public notebooks that have been shared and you'll see examples of (at least), DNNs, LSTMs & Transformer networks.\n\nThe input shape is `(None, 543, 3)` (where the initial `None` is a placeholder for a variable number of frames, the 543 is the landmarks and the 3 is the x/y/z components).",
      "votes": null
    },
    {
      "id": "2173445",
      "postDate": "03/08/2023 12:06:35",
      "content": "<p>ok,thank you so much!</p>",
      "rawMarkdown": "ok,thank you so much!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2171532,
      "author_name": "andrewrrose",
      "author_url": "",
      "post_date": "03/06/2023 21:31:19",
      "content": "<p>There's nothing that says you have to build a model a from scratch, but that's what all the top-scoring notebooks have done so far. <br>\n There are various different architectures you might want to use.  Take a look at some of the public notebooks that have been shared and you'll see examples of (at least), DNNs, LSTMs &amp; Transformer networks.</p>\n<p>The input shape is <code>(None, 543, 3)</code> (where the initial <code>None</code> is a placeholder for a variable number of frames, the 543 is the landmarks and the 3 is the x/y/z components).</p>",
      "votes": null,
      "replies": [
        {
          "id": 2173445,
          "author_name": "fatihaben",
          "author_url": "",
          "post_date": "03/08/2023 12:06:35",
          "content": "<p>ok,thank you so much!</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2171282": "Hi everyone  \n\nI want to know if we must build the model from scratch ,also if you have some propostion of what architecture to use ?\nand for the input shape is it equal to [543,3] or I am wrong ?\n\nthanks for your help",
    "2171532": "There's nothing that says you have to build a model a from scratch, but that's what all the top-scoring notebooks have done so far. \n There are various different architectures you might want to use.  Take a look at some of the public notebooks that have been shared and you'll see examples of (at least), DNNs, LSTMs & Transformer networks.\n\nThe input shape is `(None, 543, 3)` (where the initial `None` is a placeholder for a variable number of frames, the 543 is the landmarks and the 3 is the x/y/z components).",
    "2173445": "ok,thank you so much!"
  },
  "source": "meta"
}