AI Paper Atlas

Research map / 3D vision / 3D generation

SurroundOcc: Multi-camera 3D Occupancy Prediction for Autonomous Driving

Yi Wei, Linqing Zhao, Wenzhao Zheng, Zheng Zhu, Jie Zhou, Jiwen Lu

ICCV 2023 · 3D vision › 3D generation · approach: Vision-based 3D occupancy prediction · 332 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
supervise with dense occupancy labels instead of sparse LiDAR points (ablation)SurroundOcc · nuScenes valSSC mIoU: 12.17 -> 20.30Improved
densify multi-frame sparse occupancy labels with Poisson reconstruction and NN (ablation)SurroundOcc labels · nuScenes valSSC mIoU: 18.83 -> 20.30Improved
use 3D volume cross-view attention instead of averaging multi-camera features (ablation)SurroundOcc · nuScenes valSSC mIoU: 17.34 -> 20.30Improved
use 3D volume queries instead of BEV queries in cross-view attention (ablation)SurroundOcc · nuScenes valSSC mIoU: 18.94 -> 20.30Improved
add multi-scale skip connection structure (ablation)SurroundOcc · nuScenes valSSC mIoU: 18.22 -> 20.30Improved
add multi-scale supervision with decayed loss weights (ablation)SurroundOcc · nuScenes valSSC mIoU: 19.73 -> 20.30Improved
train TPVFormer with dense occupancy labels instead of sparse LiDAR pointsTPVFormer · nuScenes valSC IoU / SSC mIoU: SC IoU 11.51 -> 30.86; SSC mIoU 11.66 -> 17.10Improved
SurroundOcc applied to monocular SemanticKITTI without monocular-specific designSurroundOcc · SemanticKITTI testSSC mIoU: 11.26 -> 11.86 (vs TPVFormer)Improved
SurroundOcc versus BEVFormer in efficiencySurroundOcc · nuScenes inference, RTX 3090latency/memory: 0.31 -> 0.34 s; 4.5 -> 5.9 GHurt

Later papers that compared with this method

16 later accepted papers report SurroundOcc as a baseline in their results tables; 14 of them report a better number on most of the results they share with it.

8 more not listed.

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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