AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Unstructured Feature Decoupling for Vehicle Re-identification

Wen Qian, Hao Luo, Silong Peng, Fan Wang, Chen Chen, Hao Li

ECCV 2022 · Label-efficient and robust learning › Recognition and long-tail · approach: Cross-modality and occlusion re-identification · 44 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 transformer-based decomposing head (TDH) (ablation)ResNet50 baseline · VeRi-776mAP: 79.6 -> 81.0Improved
add transformer-based decomposing head (TDH) (ablation)Swin-Tiny baseline · VeRi-776mAP: 78.2 -> 80.0Improved
add cluster-based decoupling constraint (CDC) on top of TDH (ablation)UFDN Res50 · VeRi-776mAP: 81.0 -> 81.5Improved
add cluster-based decoupling constraint (CDC) on top of TDH (ablation)UFDN Swin-Tiny · VeRi-776mAP: 80.0 -> 80.9Improved
add diversity constraint (ablation)UFDN Res50 · VeRi-776mAP: 81.0 -> 81.2Improved
add plain alignment constraint to diversity constraint (ablation)UFDN Swin-Tiny · VeRi-776mAP: 80.5 -> 80.3Hurt
use modulated alignment constraint instead of plain one, with diversity (ablation)UFDN Res50 · VeRi-776mAP: 81.2 -> 81.5Improved
use modulated alignment constraint instead of plain one, with diversity (ablation)UFDN Swin-Tiny · VeRi-776mAP: 80.3 -> 80.9Improved
set TDH depth to 2 instead of 1 or 4 (ablation)UFDN Res50 · VehicleIDCMC@1: 86.5 -> 88.4Improved
set TDH depth to 2 instead of 4 (ablation)UFDN Res50 · VehicleIDCMC@1: 85.2 -> 88.4Improved
decouple into four parts instead of one (ablation)UFDN Res50 · VehicleIDCMC@1: 86.5 -> 88.4Improved
use 16 decoupled parts instead of 4 (ablation)UFDN Res50 · VeRi-776mAP: 81.5 -> 80.3Hurt

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