AI Paper Atlas

Research map / 3D vision / 3D generation

Structured Bird's-Eye-View Traffic Scene Understanding From Onboard Images

Yigit Baran Can, Alexander Liniger, Danda Pani Paudel, Luc Van Gool

ICCV 2021 · 3D vision › 3D generation · approach: 3D perception for autonomous driving · 122 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
add the refinement net to the object branch (ablation)STSU object branch · nuScenes objectsmIOU: 5.7 pointsImproved
add split positional encoding to the large model (ablation)STSU (Large) · nuScenes objectsobj-mean mIOU: 8.5 -> 8.7Improved
add split positional encoding to the large model (ablation)STSU (Large) · nuScenes lane graphDetect: 58.8 -> 52.8Hurt
add log of BEV locations to split positional encoding (large model) (ablation)STSU (Large + Split) · nuScenes objectsobj-mean mIOU: 8.7 -> 9.2Improved
add log of BEV locations to split positional encoding (large model) (ablation)STSU (Large + Split) · nuScenes lane graphCon-IOU: 40.8 -> 38.9Hurt
add log of BEV locations to split positional encoding (large model) (ablation)STSU (Large + Split) · nuScenes lane graphDetect: 52.8 -> 60.6Improved
add log of BEV locations to split positional encoding (small model) (ablation)STSU (Small + Split) · nuScenes objectsobj-mean mIOU: 8.9 -> 9.9Improved
add log of BEV locations to split positional encoding (small model) (ablation)STSU (Small + Split) · nuScenes lane graphCon-IOU: 41.3 -> 37.8Hurt
use the large transformer (four encoder and four decoder layers) instead of the small one, with split log (ablation)STSU · nuScenes objectsobj-mean mIOU: 9.9 -> 9.2Hurt
use the large transformer instead of the small one, plain positional embedding (ablation)STSU · nuScenes objectsobj-mean mIOU: 9.1 -> 8.5No clear effect
predict directed graph with association head instead of Polygon-RNN with estimated initial pointsSTSU vs Poly(Est) · nuScenes lane graphC-Rec: 16.3 -> 52.2Improved
use the proposed transformer instead of PON for objectsSTSU · nuScenes objectscar mIOU: 24.7 -> 32.5Improved

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