Pinned
Thrilled to share our #ICCV2023 paper: "BEVBert: Multimodal Map Pre-training for Language-guided Navigation"!
We discard traditional panoramas, instead introducing cutting-edge map-based V&L pre-training.
Pushing VLN-CE towards 60% SR for 1st time! 🥳
arxiv.org/pdf/2212.04385…
