search

Owen-Liuyuxuan / visualDet3D

Official Repo for Ground-aware Monocular 3D Object Detection for Autonomous Driving / YOLOStereo3D: A Step Back to 2D for Efficient Stereo 3D Detection
https://owen-liuyuxuan.github.io/papers_reading_sharing.github.io/3dDetection/GroundAwareConvultion/
Apache License 2.0
361 stars 76 forks source link

undefined variables #38

Closed mrsempress closed 2 years ago

mrsempress commented 2 years ago

In detection_3d_head.py line 468 and 469, you use undefined variables: anchors_3d_mean_std; I guess it should be anchor_mean_std_3d_j

Owen-Liuyuxuan commented 2 years ago

https://github.com/Owen-Liuyuxuan/visualDet3D/blob/master/visualDet3D/networks/heads/detection_3d_head.py#L467

This section is actually not tested. It should be changed to "anchor_mean_std_3d_j[pos_inds]" to only select the positive anchors, also there will be a mask output from the self._decode method.

I will do some simple test.

Owen-Liuyuxuan commented 2 years ago

I tested the code on a debug split this time. Of course the configuration files needs to be modified as well (once decoded, the number of channels will decrease, you may need to modify two lines in the config file).

Also you may want to further tune the parameters and learning rate.

Feel free to re-open the issue