BEVFormer icon indicating copy to clipboard operation
BEVFormer copied to clipboard

Published 20 hours ago • verified publisher icon fundamentalvision

Indexes for batch_size >= 2 in spatial_cross_attention.py

Open chenchaoxu opened this issue 3 years ago • 2 comments

https://github.com/zhiqi-li/BEVFormer/blob/df4023928133d3aa56c96c8185b66f15511bd86e/projects/mmdet3d_plugin/bevformer/modules/spatial_cross_attention.py#L137-L142 It seems to generate indexes for each batch per camera? Add an extra For-Loop?

chenchaoxu avatar Aug 09 '22 05:08 chenchaoxu

https://github.com/zhiqi-li/BEVFormer/blob/df4023928133d3aa56c96c8185b66f15511bd86e/projects/mmdet3d_plugin/bevformer/modules/spatial_cross_attention.py#L137-L142 It seems to generate indexes for each batch per camera? Add an extra For-Loop?

same questions, any idea or explanation for these lines? which seems like a bug? @whai362 @zhiqi-li @daijifeng001

usherbob avatar Dec 06 '23 09:12 usherbob

https://github.com/zhiqi-li/BEVFormer/blob/df4023928133d3aa56c96c8185b66f15511bd86e/projects/mmdet3d_plugin/bevformer/modules/spatial_cross_attention.py#L137-L142 It seems to generate indexes for each batch per camera? Add an extra For-Loop?

same questions, any idea or explanation for these lines? which seems like a bug? @whai362 @zhiqi-li @daijifeng001

I got same question with you. Do you solve it? Can you explain the function that you solve this? Please and Thanks!

admyxs avatar Apr 01 '24 12:04 admyxs