LSSTransform vs. BEVFormerEncoder for Image-to-BEV #205
Description
I noticed that BEVFormer uses BEVFormerEncoder for image-to-BEV transformation, while your work uses LSSTransform. I didn’t find a direct explanation or ablation in the paper—could you briefly clarify the reason for this design choice?