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
362 stars 77 forks source link

backbone weights #19

Closed vobecant closed 3 years ago

vobecant commented 3 years ago

Dear authors,

thank you very much for your repository. I would like to ask whether you use the ResNet-101 backbone that is initialized from ImageNet or from scratch?

Thank you very much in advance.

Owen-Liuyuxuan commented 3 years ago

As in code, by default, it is initialized from ImageNet. You can also toggle the flag in line 116 of the config file to do some more experiments:

https://github.com/Owen-Liuyuxuan/visualDet3D/blob/master/config/Yolo3D_example#L116

However, a performance drop is expected (my own experiments show that the performance drop is quite large, maybe I need to train for much more epochs)