search

mit-han-lab / bevfusion

[ICRA'23] BEVFusion: Multi-Task Multi-Sensor Fusion with Unified Bird's-Eye View Representation
https://bevfusion.mit.edu
Apache License 2.0
2.27k stars 409 forks source link

BEV Segmentation Map Scaling #439

Closed MaekTec closed 2 months ago

MaekTec commented 1 year ago

Hi, I saw that you are using the GlobalRotScaleTrans data augmentation, but why do you not scale the BEV ground truth map in the LoadBEVSegmentation class as well? Is it on purpose or did you just forget it? I think without the scaling, the features for the object detection are not 100% correctly aligned with the features for the map segmentation.

quotation2520 commented 3 months ago

I was wondering the same thing, and I suppose LoadBEVSegmentation do the alignment by referencing "lidar_aug_matrix" via lidar2global and map_pose. If you can look into it yourself and give confirmation, it would be very helpful.

zhijian-liu commented 2 months ago

Thank you for your interest in our project. This repository is no longer actively maintained, so we will be closing this issue. Please refer to the amazing implementation at MMDetection3D. Thank you again!