TRAILab / CaDDN

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

The inference results are 0. #82

Closed rockywind closed 2 years ago

rockywind commented 2 years ago

Hi, When I train the model in a training set that contains 3712 images and saves the models from 1 epoch to 80 epochs, But some models' results are 0. image

codyreading commented 2 years ago

Is this the KITTI dataset? And did you make any changes to the code? I did not receive this issue when I was training on KITTI.

codyreading commented 2 years ago

Closing due to inactivity.