search

alexklwong / unsupervised-depth-completion-visual-inertial-odometry

Tensorflow and PyTorch implementation of Unsupervised Depth Completion from Visual Inertial Odometry (in RA-L January 2020 & ICRA 2020)
Other
192 stars 24 forks source link

How to generate colored and back-projected to 3-D? #4

Closed chl2 closed 1 year ago

chl2 commented 2 years ago

Hi, how did you generate the predicted depth map of the kitti dataset as colored and back-projected to 3-D? I think the net_utils file seems to be doing this, but how is it called in kitti? Thank you very much for any reply.

alexklwong commented 2 years ago

You can follow the general instructions in this thread of our recent work KBNet: https://github.com/alexklwong/calibrated-backprojection-network/issues/17#issuecomment-1176431551

This repo is in Tensorflow so it is a bit harder to use for the general purpose of creating point clouds. A PyTorch version is coming soon.

In the meantime, here is the PyTorch version of the function: https://github.com/alexklwong/calibrated-backprojection-network/blob/master/src/net_utils.py#L1638-L1667