{
  "id": 178974,
  "title": "Data Shape Intuition",
  "url": "/competitions/lyft-motion-prediction-autonomous-vehicles/discussion/178974",
  "author_name": "",
  "post_date": "2020-09-01T03:33:36.677330300Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Trying to wrap my head around this.  In the config files, there is [history_num_frames] parameter.  The data history positions/yaws have shape (history_num_frames+1, 2), which makes sense given the current frame + how ever many before.</p>\n<p>The data['image'] has (2*history_num_frames + 3 images).  I cannot find in the papers or documentation where that number comes from and what the interpretation of each channel is.  I was hoping for some clarification on what the interpretation of that first axis is.</p>",
  "messages": [
    {
      "id": "993602",
      "postDate": "09/01/2020 03:33:36",
      "content": "<p>Trying to wrap my head around this.  In the config files, there is [history_num_frames] parameter.  The data history positions/yaws have shape (history_num_frames+1, 2), which makes sense given the current frame + how ever many before.</p>\n<p>The data['image'] has (2*history_num_frames + 3 images).  I cannot find in the papers or documentation where that number comes from and what the interpretation of each channel is.  I was hoping for some clarification on what the interpretation of that first axis is.</p>",
      "rawMarkdown": "Trying to wrap my head around this.  In the config files, there is [history_num_frames] parameter.  The data history positions/yaws have shape (history_num_frames+1, 2), which makes sense given the current frame + how ever many before.\n\nThe data['image'] has (2*history_num_frames + 3 images).  I cannot find in the papers or documentation where that number comes from and what the interpretation of each channel is.  I was hoping for some clarification on what the interpretation of that first axis is.",
      "votes": null
    },
    {
      "id": "993688",
      "postDate": "09/01/2020 05:31:02",
      "content": "<p>Hi, a similar topic was discussed <a href=\"https://www.kaggle.com/c/lyft-motion-prediction-autonomous-vehicles/discussion/178097\" target=\"_blank\">here</a></p>",
      "rawMarkdown": "Hi, a similar topic was discussed [here](https://www.kaggle.com/c/lyft-motion-prediction-autonomous-vehicles/discussion/178097)",
      "votes": null
    },
    {
      "id": "994491",
      "postDate": "09/01/2020 16:05:13",
      "content": "<p>So +3 are for RGB channels and for 2*(history_num_frames) if history frames are 1 then 1 channel is for location of Ego on the image and second channel is for location of other agents on the image. So it depends on how many history frames you choose.</p>",
      "rawMarkdown": "So +3 are for RGB channels and for 2*(history_num_frames) if history frames are 1 then 1 channel is for location of Ego on the image and second channel is for location of other agents on the image. So it depends on how many history frames you choose.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 993688,
      "author_name": "micajoumathematics",
      "author_url": "",
      "post_date": "09/01/2020 05:31:02",
      "content": "<p>Hi, a similar topic was discussed <a href=\"https://www.kaggle.com/c/lyft-motion-prediction-autonomous-vehicles/discussion/178097\" target=\"_blank\">here</a></p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 994491,
      "author_name": "sujaydk",
      "author_url": "",
      "post_date": "09/01/2020 16:05:13",
      "content": "<p>So +3 are for RGB channels and for 2*(history_num_frames) if history frames are 1 then 1 channel is for location of Ego on the image and second channel is for location of other agents on the image. So it depends on how many history frames you choose.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "993602": "Trying to wrap my head around this.  In the config files, there is [history_num_frames] parameter.  The data history positions/yaws have shape (history_num_frames+1, 2), which makes sense given the current frame + how ever many before.\n\nThe data['image'] has (2*history_num_frames + 3 images).  I cannot find in the papers or documentation where that number comes from and what the interpretation of each channel is.  I was hoping for some clarification on what the interpretation of that first axis is.",
    "993688": "Hi, a similar topic was discussed [here](https://www.kaggle.com/c/lyft-motion-prediction-autonomous-vehicles/discussion/178097)",
    "994491": "So +3 are for RGB channels and for 2*(history_num_frames) if history frames are 1 then 1 channel is for location of Ego on the image and second channel is for location of other agents on the image. So it depends on how many history frames you choose."
  },
  "source": "meta"
}