AI Paper Atlas

Research map / Detection and segmentation / CNN features

Discriminative Feature Transformation for Occluded Pedestrian Detection

Chunluan Zhou, Ming Yang, Junsong Yuan

ICCV 2019 · Detection and segmentation › CNN features · approach: Pedestrian and crowd detection · 60 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 attention module to Fast R-CNNFRCN · Caltech Heavylog-average miss rate: 44.3 -> 39.7Improved
add attention module to Fast R-CNNFRCN · CityPersons Heavylog-average miss rate: 49.2 -> 47.4Improved
add negative-only feature transformation (NT) (ablation)FRCN+A · Caltech Reasonablelog-average miss rate: 8.8 -> 8.5No clear effect
add positive-only feature transformation (PT) (ablation)FRCN+A · Caltech Partiallog-average miss rate: 15.1 -> 13.1Improved
add discriminative transformation (DT) (ablation)FRCN+A · Caltech Reasonablelog-average miss rate: 8.8 -> 8.0Improved
add discriminative transformation (DT) (ablation)FRCN+A · Caltech Partiallog-average miss rate: 15.1 -> 12.2Improved
add discriminative transformation (DT) (ablation)FRCN+A · Caltech Heavylog-average miss rate: 39.7 -> 37.9Improved
add feature transformation loss to transform branch (ablation)FRCN+A+TB · Caltech Partiallog-average miss rate: 14.4 -> 12.2Improved
add discriminative transformation (DT) (ablation)FRCN+A · CityPersons Heavylog-average miss rate: 47.4 -> 44.3Improved
add transformation loss over transform branch alone (ablation)FRCN+A+TB · CityPersons Heavylog-average miss rate: 46.6 -> 44.3Improved
vary overlap thresholds tau1/tau2 for centroids (ablation)FRCN+A+DT · Caltechlog-average miss rate: 8.0 +/- 0.4 / 12.2 +/- 0.6 / 37.9 +/- 0.5No clear effect
vary score thresholds s1/s2 for centroids (ablation)FRCN+A+DT · Caltechlog-average miss rate: 8.0 +/- 0.5 / 12.2 +/- 0.7 / 37.9 +/- 0.5No clear effect

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