Depth information

[LiliMeng]

For the training, only the camera depth information is needed, or the depth information for every pixel in the image is also necessary? As it seems that the ground truth of RGBD benchmark only includes the 6D camera pose rather than the depth information for every image pixel.

Thanks a lot!

[connerbrooks]

For training you need RGB images, depth images, and camera poses. The depth information for each pixel is in the depth images. The depth values are projected into camera space and the poses are used to transform these depth values from camera space to world coordinates.