search

hustvl / GKT

Efficient and Robust 2D-to-BEV Representation Learning via Geometry-guided Kernel Transformer
https://arxiv.org/abs/2206.04584
MIT License
218 stars 18 forks source link

Can not reproduce the comparative results with BEVFormer #8

Open Zixin-Lei opened 1 year ago

Zixin-Lei commented 1 year ago

Hi! Thanks a lot for the open-source project! However, I failed to reproduce the experimental results, i.e. the high FPS compared with BEVFormer. May I ask if you have made the size of BEV query the same when doing comparison with BEVFormer? It would be great if you could share your code of the comparative experiment between GKT and BEVFormer.