Camera depth error when use Ray Tracing

[ZanxinChen]

I found that when ray tracing is enabled in the SAPIEN environment (i.e., using sapien.render.set_camera_shader_dir("rt")), I am unable to successfully filter out pixels at infinity when retrieving the depth map from the camera.

Specifically, I can't use the under code to filter out points beyond the far plane of the camera frustum:

position = camera.get_picture("Position")
position[..., 3] == 1

All values of position[..., 3] are less than 1.

Is there any solution to this issue? Thank you very much.

[fbxiang]

Currently ray tracing does not put anything in the w channel of the position texture. But the xyz channels of the background pixels should all be 0 so you can filter it out this way. A potentially safer approach is to filter by segmentation where the segmentation values of background are always 0.