how can I convert the depth map into a point cloud, given the accurate camera intrinsics

[elenacliu]

[SwcK423]

Like the depth map in lego's test folder in the synthetic dataset, which is in png format, what do we do to align it with the nerf coordinates, expecting answer.

[GauravNerf]

depth map in the test folder is not used in the training of the model. it is kept for the user to compare the results after training with gt depth. technically i see the depth map is aligned for every rgb image, so the coordinates of the rgb image may also be the same for depth images, check the transform_test.json