search

ADLab-AutoDrive / BEVHeight

An official code release of our CVPR'23 paper, BEVHeight
MIT License
195 stars 25 forks source link

Have you attempt to use auxiliary loss about the height liking BevDepth, what about the results, no improvement? #22

Closed mayechi closed 1 year ago

yanglei18 commented 1 year ago

Please refer to the supplementary materials of our CVPR paper.

mayechi commented 1 year ago

I can not find it, which chapter said the above in the materials of your paper?

yanglei18 commented 1 year ago

please refer to https://openaccess.thecvf.com/content/CVPR2023/supplemental/Yang_BEVHeight_A_Robust_CVPR_2023_supplemental.pdf

mayechi commented 1 year ago

I saw it , thank you. But I saw AP with ped and cyc in the table 5 are much lower than in the table 1 on the main paper, why? Some differences about metric over these tables?