AI Paper Atlas

Research map / Detection and segmentation / 3D detection and driving

Revisiting Token Compression for Accelerating ViT-based Sparse Multi-View 3D Object Detectors

Mingqian Ji, Shanshan Zhang, Jian Yang

CVPR 2026 · Detection and segmentation › 3D detection and driving · approach: Unified LiDAR 3D detection · 0 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 SPSS adaptive patch size (ablation)StreamPETR ViT-L · nuScenes valNDS / inference time: 61.2 -> 58.8 NDS; 317.0 -> 189.1 msMixed
add CGFE after SPSS (ablation)SPSS-enabled StreamPETR · nuScenes valNDS: 58.8 -> 60.4Improved
add IPS after SPSS and CGFE (ablation)SPSS+CGFE StreamPETR · nuScenes valinference time: 209.6 -> 194.3 msImproved
use adaptive (18, 20) patch size instead of fixed 20 (ablation)SEPatch3D · nuScenes valNDS: 59.3 -> 60.3Improved
use adaptive (18, 20) patch size instead of fixed 18 (ablation)SEPatch3D · nuScenes valinference time: 242.6 -> 194.3 msImproved
use entropy-based selection instead of score-based (ablation)IPS module · nuScenes valNDS: 59.8 -> 60.3Improved
use entropy-based selection instead of attention-based (ablation)IPS module · nuScenes valNDS: 59.6 -> 60.3Improved
use adaptive selection instead of Top-K 30% (ablation)IPS module · nuScenes valNDS: 59.6 -> 60.3Improved
use spatial cue only instead of spatial and temporal cues (ablation)SPSS · nuScenes valNDS: 60.3 -> 59.7Hurt
use temporal cue only instead of spatial and temporal cues (ablation)SPSS · nuScenes valNDS: 60.3 -> 59.5Hurt
apply SEPatch3D instead of Cropr token pruningStreamPETR · nuScenes valmFP: 53.2% -> 40.5%Improved
apply SEPatch3D-faster to DETR3DDETR3D ViT-L · nuScenes valinference time: 344.9 -> 216.2 msImproved

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