TempBEV: Improving Learned BEV Encoders with Combined Image and BEV Space Temporal Aggregation

arXiv:2404.11803v1 Announce Type: cross
Abstract: Autonomous driving requires an accurate representation of the environment. A strategy toward high accuracy is to fuse data from several sensors. Learned Bird's-Eye View (BEV) encoders can achieve this by mapping data from individual sensors into one joint latent space. For cost-efficient camera-only systems, this provides an effective mechanism to fuse data from multiple cameras with different views. Accuracy can further be improved by aggregating sensor information over time. This is especially important in monocular …

abstract accuracy aggregation arxiv autonomous autonomous driving bird cost cs.ai cs.cv cs.lg cs.ro data driving environment image improving mapping representation sensors space strategy temporal the environment type view