{
  "id": 585791,
  "title": "TOF Sensor Orientation?",
  "url": "/competitions/cmi-detect-behavior-with-sensor-data/discussion/585791",
  "author_name": "",
  "post_date": "2025-06-23T06:26:32.683211Z",
  "votes": 4,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Was wondering if anyone can confirm if preprocessing was done on the TOF sensor input and if so what preprocessing. Some are in a horizontal orientation and others vertical and the datasheet also mentions flipped input across horizontal and vertical axes. </p>\n<p>Wondering if all inputs were transposed accordingly such that they followed the data tab's 0,1,2,3,4,5,6,7 orientation or if they vary based on their sensor orientation on the board. I'm hoping to orient the data such that they are all representing the same view just across the 5 inputs so wanted to confirm if this was already handled.</p>\n<p>The left and right TOF sensors appear to have the RX side on the upper side while all other three have it on the right side.</p>\n<p>Thank you.</p>",
  "messages": [
    {
      "id": "3230520",
      "postDate": "06/23/2025 06:26:32",
      "content": "<p>Was wondering if anyone can confirm if preprocessing was done on the TOF sensor input and if so what preprocessing. Some are in a horizontal orientation and others vertical and the datasheet also mentions flipped input across horizontal and vertical axes. </p>\n<p>Wondering if all inputs were transposed accordingly such that they followed the data tab's 0,1,2,3,4,5,6,7 orientation or if they vary based on their sensor orientation on the board. I'm hoping to orient the data such that they are all representing the same view just across the 5 inputs so wanted to confirm if this was already handled.</p>\n<p>The left and right TOF sensors appear to have the RX side on the upper side while all other three have it on the right side.</p>\n<p>Thank you.</p>",
      "rawMarkdown": "Was wondering if anyone can confirm if preprocessing was done on the TOF sensor input and if so what preprocessing. Some are in a horizontal orientation and others vertical and the datasheet also mentions flipped input across horizontal and vertical axes. \n\nWondering if all inputs were transposed accordingly such that they followed the data tab's 0,1,2,3,4,5,6,7 orientation or if they vary based on their sensor orientation on the board. I'm hoping to orient the data such that they are all representing the same view just across the 5 inputs so wanted to confirm if this was already handled.\n\nThe left and right TOF sensors appear to have the RX side on the upper side while all other three have it on the right side.\n\nThank you.",
      "votes": null
    },
    {
      "id": "3231360",
      "postDate": "06/24/2025 08:12:39",
      "content": "<p>For me, I think we can train a Orientation Classifier to predict the orientation of test dataset, at the same time, we add the 'Orientation'  as  our training-features (our <strong>three_branch_model</strong> will use it). At the test stage, we first use the Orientation Classifier to predict the orientation of test sample then, we can use our <strong>three_branch_model</strong> to get the result on it. <br>\nI hope my idea may be helpful for you.  : )</p>",
      "rawMarkdown": "For me, I think we can train a Orientation Classifier to predict the orientation of test dataset, at the same time, we add the 'Orientation'  as  our training-features (our **three_branch_model** will use it). At the test stage, we first use the Orientation Classifier to predict the orientation of test sample then, we can use our **three_branch_model** to get the result on it. \nI hope my idea may be helpful for you.  : )",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3231360,
      "author_name": "lau01b",
      "author_url": "",
      "post_date": "06/24/2025 08:12:39",
      "content": "<p>For me, I think we can train a Orientation Classifier to predict the orientation of test dataset, at the same time, we add the 'Orientation'  as  our training-features (our <strong>three_branch_model</strong> will use it). At the test stage, we first use the Orientation Classifier to predict the orientation of test sample then, we can use our <strong>three_branch_model</strong> to get the result on it. <br>\nI hope my idea may be helpful for you.  : )</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3230520": "Was wondering if anyone can confirm if preprocessing was done on the TOF sensor input and if so what preprocessing. Some are in a horizontal orientation and others vertical and the datasheet also mentions flipped input across horizontal and vertical axes. \n\nWondering if all inputs were transposed accordingly such that they followed the data tab's 0,1,2,3,4,5,6,7 orientation or if they vary based on their sensor orientation on the board. I'm hoping to orient the data such that they are all representing the same view just across the 5 inputs so wanted to confirm if this was already handled.\n\nThe left and right TOF sensors appear to have the RX side on the upper side while all other three have it on the right side.\n\nThank you.",
    "3231360": "For me, I think we can train a Orientation Classifier to predict the orientation of test dataset, at the same time, we add the 'Orientation'  as  our training-features (our **three_branch_model** will use it). At the test stage, we first use the Orientation Classifier to predict the orientation of test sample then, we can use our **three_branch_model** to get the result on it. \nI hope my idea may be helpful for you.  : )"
  },
  "source": "meta"
}