search

TRAILab / CaDDN

Categorical Depth Distribution Network for Monocular 3D Object Detection (CVPR 2021 Oral)
Apache License 2.0
359 stars 62 forks source link

Generation of labels on depth_2 file #43

Closed YunzheWu-404 closed 3 years ago

YunzheWu-404 commented 3 years ago

Hi

Thank you very much for sharing your great work with us.

I am a big fun of your use of semantic segmentation and pcdet in 3D detection.

Could I ask detatils about how you trasform the sparse depth labels provided in kitti to the denth labels in depth_2?

I would really appreciate if you could help.

Best.

codyreading commented 3 years ago

Hi and thanks for the interest!

For the KITTI Dataset, I used the depth completion method IP-Basic to generate dense depth maps from LiDAR data. You could do the same for another dataset, or alternatively use a more current depth completion method to do the same thing. I used the following parameters on the KITTI dataset:

fill_type = 'multiscale'
extrapolate = False
blur_type = 'bilateral'