Research map / Label-efficient and robust learning / Semi-supervised learning
Unlabeled Samples Generated by GAN Improve the Person Re-Identification Baseline in Vitro
Zhedong Zheng, Liang Zheng, Yi Yang
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add 24,000 GAN images with LSRO | ResNet-50 baseline · Market-1501 | rank-1 accuracy (%): 73.69 -> 78.06 | Improved |
| add 24,000 GAN images with LSRO | ResNet-50 baseline · Market-1501 | mAP (%): 51.48 -> 56.23 | Improved |
| add GAN images with LSRO | ResNet-50 baseline · DukeMTMC-reID | rank-1 accuracy (%): 65.22 -> 67.68 | Improved |
| add GAN images (from Market-1501) with LSRO | ResNet-50 baseline · CUHK03 | rank-1 accuracy (%): 71.5 -> 73.1 | Improved |
| add GAN images with LSRO | Verif-Identif. model · Market-1501 | rank-1 accuracy (%): 79.51 -> 83.97 | Improved |
| add GAN images with LSRO | ResNet-50 baseline · CUB-200-2011 | top-1 accuracy (%): 82.6 -> 83.2 | Improved |
| increase GAN images from 24,000 to 36,000 (ablation) | LSRO · Market-1501 | rank-1 accuracy (%): 78.06 -> 76.07 | Hurt |
| use 12,000 instead of 24,000 GAN images (ablation) | LSRO · Market-1501 | rank-1 accuracy (%): 78.06 -> 76.81 | Hurt |
| use LSRO instead of All in one (ablation) | GAN-augmented training · Market-1501 (24,000 GAN images) | rank-1 accuracy (%): 76.63 -> 78.06 | Improved |
| use LSRO instead of Pseudo label (ablation) | GAN-augmented training · Market-1501 (24,000 GAN images) | rank-1 accuracy (%): 75.80 -> 78.06 | Improved |
| use GAN images instead of real CUHK03 images as unlabeled data (ablation) | LSRO · Market-1501 (12,000 unlabeled images) | rank-1 accuracy (%): 75.65 -> 76.81 | Improved |
| add real CUHK03 images as unlabeled data (ablation) | LSRO · Market-1501 | rank-1 accuracy (%): 73.69 -> 75.65 | Improved |
Closest papers by meaning
- Pose-Normalized Image Generation for Person Re-identification (ECCV 2018 · similarity 0.89)
- Learning Invariance From Generated Variance for Unsupervised Person Re-Identification (TPAMI 2022 · similarity 0.88)
- Joint Generative and Contrastive Learning for Unsupervised Person Re-Identification (CVPR 2021 · similarity 0.88)
- Learning Disentangled Representation for Robust Person Re-identification (NeurIPS 2019 · similarity 0.88)
- FD-GAN: Pose-guided Feature Distilling GAN for Robust Person Re-identification (NeurIPS 2018 · similarity 0.87)
- Adversarially Occluded Samples for Person Re-Identification (CVPR 2018 · similarity 0.87)
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.
