different methods to obtain ref_local_d

[Valentina-Zhang]

Hello, I noticed that there are two different approaches to obtain ref_local_d in the train.py file:

ref_local_d = render_pkg['rendered_distance'].squeeze()
# rays_d = viewpoint_cam.get_rays()
# rendered_normal2 = rendered_normal.reshape(-1,3)
# ref_local_d = render_pkg['plane_depth'].view(-1) * ((rendered_normal2 * rays_d.reshape(-1,3)).sum(-1).abs())
# ref_local_d = ref_local_d.reshape(H,W)

I was wondering if there are any significant differences between the current method and the commented-out code (apart from potential CUDA acceleration)? Do they produce identical results?

Thank you for your time and for sharing this great work!

[xmlyqing00]

I computed the difference between these depth maps, they are different LOL