AI Paper Atlas

Research map / 3D vision / 3D generation

Accelerating Online Mapping and Behavior Prediction via Direct BEV Feature Attention

Xunjiang Gu, Guanyu Song, Igor Gilitschenski, Marco Pavone, Boris Ivanovic

ECCV 2024 · 3D vision › 3D generation · approach: 3D perception for autonomous driving · 10 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
feed BEV features directly to the prediction model instead of decoded vector mapsHiVT with MapTR · nuScenesminFDE: 0.8900 -> 0.7401Improved
feed BEV features to DenseTNT instead of vector mapsDenseTNT with MapTR · nuScenesminFDE: 2.0661 -> 1.4700Improved
add BEV features when the map model also outputs centerlinesMapTRv2-CL with HiVT/DenseTNT · nuScenesminFDE: DenseTNT 1.3174 -> 1.3609No clear effect
add uncertainty-enhanced vertices (prior work) for DenseTNT with MapTRDenseTNT · nuScenesminFDE: 2.0661 -> 2.1502Hurt
skip map decoding and attend to BEV features directlyHiVT with MapTR · nuScenes inferenceruntime (ms): 22.4 -> 9.1Improved
skip map decoding with StreamMapNetHiVT with StreamMapNet · nuScenes inferenceruntime (ms): 33.6 -> 29.4Improved
use BEV patch 10x5 instead of 20x20 (ablation)HiVT + MapTRv2-Centerline · nuScenesminFDE: 0.7518 -> 0.8003Hurt
use larger BEV patch 40x20 instead of 20x20 (ablation)HiVT + MapTRv2-Centerline · nuScenesMR: 0.0737 -> 0.0770Hurt
use temporal BEV features (BEVFormer-based MapTR, StreamMapNet) instead of single-frame LSS features (ablation)BEV feature prediction · nuScenesminADE/minFDE/MR: 16%, 19%, 24% vs 4%, 2%, 4% relative improvementImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy