search

huang-yh / SelfOcc

[CVPR 2024] SelfOcc: Self-Supervised Vision-Based 3D Occupancy Prediction
Apache License 2.0
273 stars 17 forks source link

Camera synchronisation, calibration & real-time inference #18

Open Tuhin-nd opened 1 month ago

Tuhin-nd commented 1 month ago

Dear @huang-yh / @wzzheng , Thanks for the wonderful paper and GitHub repo. Could you please help me with the below doubt?

  1. Does this method require training data from multiple cameras to be synchronised at sub-millisec level?
  2. Can this method run at an edge device similar to Nvidia Jetson for real time inference ~ 30 fps speed?
  3. For feeding the inputs do we need to undistort the images and "pre-compute" intrinsics & extrinsics of all the cameras?