AI Paper Atlas

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

Attack-Guided Perceptual Data Generation for Real-World Re-Identification

Yukun Huang, Xueyang Fu, Zheng-Jun Zha

ICCV 2021 · Label-efficient and robust learning › Recognition and long-tail · approach: Supervised person re-identification · 12 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 generated samples trained with plain cross-entropy only (ablation)ResNet-50 baseline · MLR-CUHK03Rank-1 (%): 78.1 -> 75.2Hurt
add self-center loss and Wasserstein loss on generated samples (ablation)GAAG naive · MLR-CUHK03Rank-1 (%): 75.2 -> 87.6Improved
add synthetic-to-real feature constraints (ablation)GAAG naive · MLR-VIPeRRank-1 (%): 29.7 -> 52.2Improved
add global-aware dense augmentation (ablation)GAAG · MLR-CUHK03Rank-1 (%): 84.8 -> 87.6Improved
add global-aware dense augmentation (ablation)GAAG · MLR-VIPeRRank-1 (%): 49.7 -> 52.2Improved
add global-aware dense augmentation (ablation)GAAG · CAVIARRank-1 (%): 43.8 -> 44.0No clear effect
add ReID-driven perceptual attack guidance (ablation)GAAG · MLR-CUHK03Rank-1 (%): 86.1 -> 87.6Improved
add ReID-driven perceptual attack guidance (ablation)GAAG · MLR-VIPeRRank-1 (%): 51.3 -> 52.2Improved
add ReID-driven perceptual attack guidance (ablation)GAAG · CAVIARRank-1 (%): 41.6 -> 44.0Improved
add random down-sampling augmentation with probability 0.5 (ablation)ResNet-50 baseline · MLR-CUHK03Rank-1 (%): 78.1 -> 76.3Hurt
add random down-sampling augmentation with probability 0.5 (ablation)ResNet-50 baseline · CAVIARRank-1 (%): 25.8 -> 21.4Hurt
add GAAG perceptual data generation (illumination degradation)ResNet-50 baseline · MSMT17Rank-1 (%): 64.0 -> 73.3Improved

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