search

fundamentalvision / BEVFormer

[ECCV 2022] This is the official implementation of BEVFormer, a camera-only framework for autonomous driving perception, e.g., 3D object detection and semantic map segmentation.
https://arxiv.org/abs/2203.17270
Apache License 2.0
3.4k stars 550 forks source link

What is the usage of self.rotation_pre_bev? #125

Closed Qi-Zhangyang closed 2 years ago

Qi-Zhangyang commented 2 years ago

https://github.com/fundamentalvision/BEVFormer/blob/074e3688cc4fa30dd4de3ac034227f125b61157e/projects/mmdet3d_plugin/bevformer/modules/transformer.py#L143-L156

I wanna know if rotation of pre_bev can improve the performence?

Thank u very much.

Qi-Zhangyang commented 2 years ago

And there is a small error. The rotation center is always [100, 100] whatever the bev_h and bev_w are.

https://github.com/fundamentalvision/BEVFormer/blob/074e3688cc4fa30dd4de3ac034227f125b61157e/projects/mmdet3d_plugin/bevformer/modules/transformer.py#L50-L68

zhiqi-li commented 2 years ago

And there is a small error. The rotation center is always [100, 100] whatever the bev_h and bev_w are.

https://github.com/fundamentalvision/BEVFormer/blob/074e3688cc4fa30dd4de3ac034227f125b61157e/projects/mmdet3d_plugin/bevformer/modules/transformer.py#L50-L68

Yes, while using different bev size, we should adjust the rotate_center.

Qi-Zhangyang commented 2 years ago

Thanks very much.