StructDepth
StructDepth copied to clipboard
Camera Intrinsics for pretrained model
Hi there,
Could you please tell me what camera intrinsic are used for the pretrained models? I would like to see some depth results on a custom input in the form of a pointcloud, similar to #2. It is unclear to me what intrinsics I should use for these images: the values found in nyu_dataset? and with or without crop? Also what is the purpose of dividing by height and width?
Thanks in advance for your help