assets.md

May 5, 2023 ยท View on GitHub

Note that the mIoU is silighter higher than those in our paper (see logs for more details). Bellow are several changes in the training phase:

  • We use SyncBN instead of BN/GN, which enhances the mIoU by ~0.3.
  • The total training epoch is 15 instead of 24, which significantly reduces the training time.
  • The models are trained with annotation-v0.1 (with less occupancy artifacts).
SubsetCheckpointLogsNote
Camera-based baselinelink (code:tlif)link (code:ahqs)train on 8 RTX3090
LiDAR-based baselinelink (code:qdsl)link (code:p3ra)train on 8 RTX3090
Multimodal baselinelink (code:d3vl)link (code:f5qq)train on 8 RTX3090
Camera-based CONetlink (code:630w)link (code:jb9o)train on 8 A100
LiDAR-based CONetlink (code:hnaf)link (code:hqto)train on 8 RTX3090
Multimodal CONetlink (code:k9p9)link (code:t7c5)train on 8 A100

Google Drive link