{
  "id": 190519,
  "title": "Rotation transformational inputs",
  "url": "/competitions/lyft-motion-prediction-autonomous-vehicles/discussion/190519",
  "author_name": "shashank dahiya",
  "post_date": "2020-10-12T06:19:20.673000",
  "votes": 1,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Okay, how prediction working perfectly? None of the agents has some target path to follow so each agent may go every path possibly. But the model predicts the target.<br>\nSo after spending hours in the process I think input does have some rotation transformation in itself. </p>\n<p>On visualizing output by transform target coordinates to the image, each agent has that target trajectory, it goes better. <br>\nDiscussion: [<a href=\"https://www.kaggle.com/c/lyft-motion-prediction-autonomous-vehicles/discussion/186492\" target=\"_blank\">link</a>]</p>",
  "messages": [
    {
      "id": 1046961,
      "postDate": "2020-10-12T06:19:20.673Z",
      "content": "<p>Okay, how prediction working perfectly? None of the agents has some target path to follow so each agent may go every path possibly. But the model predicts the target.<br>\nSo after spending hours in the process I think input does have some rotation transformation in itself. </p>\n<p>On visualizing output by transform target coordinates to the image, each agent has that target trajectory, it goes better. <br>\nDiscussion: [<a href=\"https://www.kaggle.com/c/lyft-motion-prediction-autonomous-vehicles/discussion/186492\" target=\"_blank\">link</a>]</p>",
      "rawMarkdown": "Okay, how prediction working perfectly? None of the agents has some target path to follow so each agent may go every path possibly. But the model predicts the target.\nSo after spending hours in the process I think input does have some rotation transformation in itself. \n\nOn visualizing output by transform target coordinates to the image, each agent has that target trajectory, it goes better. \nDiscussion: [[link](https://www.kaggle.com/c/lyft-motion-prediction-autonomous-vehicles/discussion/186492)]",
      "votes": 1
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1046961": "Okay, how prediction working perfectly? None of the agents has some target path to follow so each agent may go every path possibly. But the model predicts the target.\nSo after spending hours in the process I think input does have some rotation transformation in itself. \n\nOn visualizing output by transform target coordinates to the image, each agent has that target trajectory, it goes better. \nDiscussion: [[link](https://www.kaggle.com/c/lyft-motion-prediction-autonomous-vehicles/discussion/186492)]"
  }
}