This repository contains the code for the CVPR 2020 paper "Differentiable Volumetric Rendering: Learning Implicit 3D Representations without 3D Supervision"
Hi, I see that in your code, you get ray_vector by subtracting the camera origin (camera_world) from the pixel (p_world) ray_vector = camera_world - p_world. I'm wondering if this ray_vector can have a depth other than 1? For example, can you sample a point inside the unit cube and set ray_vector = p_inside_cube - camera_world, will this ray_vector still be a valid input into the ray marching function? Thanks!
Hi, I see that in your code, you get ray_vector by subtracting the camera origin (camera_world) from the pixel (p_world)
ray_vector = camera_world - p_world.
I'm wondering if this ray_vector can have a depth other than 1? For example, can you sample a point inside the unit cube and setray_vector = p_inside_cube - camera_world
, will this ray_vector still be a valid input into the ray marching function? Thanks!