Sorry to bother you. I want to use the ground truth depthmap as another supervision signal. I want to know the definition of rendered depthmap of S-NeRF. Does the rendered depth mean the Euclidean distance from the camera center (spherical center) to the 3D point? Or it means the z-depth like perspective images?
I ask this question because the result, after I used the ground depth map as supervision signal, is worse than that I did not use. I guess I maybe use the depth in a wrong way. (Tested in the wide-baseline of Matterport3d dataset)
Sorry to bother you. I want to use the ground truth depthmap as another supervision signal. I want to know the definition of rendered depthmap of S-NeRF. Does the rendered depth mean the Euclidean distance from the camera center (spherical center) to the 3D point? Or it means the z-depth like perspective images?
I ask this question because the result, after I used the ground depth map as supervision signal, is worse than that I did not use. I guess I maybe use the depth in a wrong way. (Tested in the wide-baseline of Matterport3d dataset)