AI Paper Atlas

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

Cross-Category Subjectivity Generalization for Style-Adaptive Sketch Re-ID

Zechao Hu, Zhengwei Yang, Hao Li, Zheng Wang, Yixiong Zou

ICCV 2025 · Label-efficient and robust learning › Recognition and long-tail · approach: Supervised person re-identification · 2 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 AIP over CLIP-base trained on MaSk1k only (ablation)CLIP-base · PKU-SketchmAP: 87.5 -> 94.3Improved
add Sketchy then fine-tune on MaSk1k instead of MaSk1k only (no AIP) (ablation)CLIP-base · PKU-SketchmAP: 87.5 -> 88.6Improved
naively mix Sketchy with MaSk1k (ablation)CLIP-base · PKU-SketchmAP: 87.5 -> 87.1Hurt
add trainable LayerNorm (ablation)AIP · PKU-SketchmAP: 89.4 -> 94.3Improved
add category-wise prompt (ablation)AIP · PKU-SketchmAP: 89.3 -> 94.3Improved
add globally shared prompt (ablation)AIP · PKU-SketchmAP: 93.0 -> 94.3Improved
add ACDE (ablation)AIP · PKU-SketchmAP: 92.7 -> 94.3Improved
add MOIA meta-optimization (ablation)AIP · PKU-SketchmAP: 92.4 -> 94.3Improved
add prompt generator phi (ablation)AIP · PKU-SketchmAP: 91.6 -> 94.3Improved
add MOIA meta-optimization (ablation)AIP · PKU-SketchRank-1: 98.0 -> 96.0Hurt
increase proportion of Sketchy non-pedestrian data to 10% (ablation)AIP · PKU-SketchmAP: 10% (about 1.2k RGB, 6k sketches)Improved
use more than 70% Sketchy data (ablation)AIP · PKU-SketchmAP: 70% (about 8.4k RGB, 42k sketches)No 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