Research map / Vision-language and multimodal / CLIP and zero-shot
Unified Pre-Training with Pseudo Texts for Text-To-Image Person Re-Identification
Zhiyin Shao, Xinyu Zhang, Changxing Ding, Jian Wang, Jingdong Wang
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 MLM objective to contrastive pre-training on LUPerson-T (ablation) | UniPT · CUHK-PEDES | Rank-1: 65.77 -> 66.83 | Improved |
| add MLM objective to contrastive pre-training on LUPerson-T (ablation) | UniPT · ICFG-PEDES | Rank-1: 57.35 -> 59.08 | Improved |
| add granularity-unified loss L_pgu during supervised fine-tuning (ablation) | UniPT · ICFG-PEDES | Rank-1: 54.41 -> 59.08 | Improved |
| unified vision-language pre-training instead of separate pre-training (ablation) | T2I-ReID · CUHK-PEDES | Rank-1: 65.30 -> 66.83 | Improved |
| pre-train on LUPerson-T pseudo-text pairs versus generic pre-training (ViT-B/16) | T2I-ReID · CUHK-PEDES | Rank-1: +1.91 | Improved |
| UniPT pre-training versus CLIP initialization with ViT-B/16 | T2I-ReID · CUHK-PEDES | Rank-1: 66.34 -> 68.50 | Improved |
| use more templates for pseudo-text generation (ablation) | LUPerson-T · CUHK-PEDES | Rank-1 | Improved |
| include more images with the same number of templates (ablation) | LUPerson-T · CUHK-PEDES | Rank-1 | Hurt |
| UniPT pre-trained model on cross-domain transfer versus SSAN (w/ BERT) | UniPT · ICFG-PEDES to CUHK-PEDES (cross-domain) | Rank-1: 21.07 -> 28.48 | Improved |
Closest papers by meaning
- Noisy-Correspondence Learning for Text-to-Image Person Re-identification (CVPR 2024 · similarity 0.88)
- Modeling Thousands of Human Annotators for Generalizable Text-to-Image Person Re-identification (CVPR 2025 · similarity 0.88)
- Multi-Prompts Learning with Cross-Modal Alignment for Attribute-Based Person Re-identification (AAAI 2024 · similarity 0.88)
- Unifying Multi-Modal Uncertainty Modeling and Semantic Alignment for Text-to-Image Person Re-identification (AAAI 2024 · similarity 0.87)
- TF-CLIP: Learning Text-Free CLIP for Video-Based Person Re-identification (AAAI 2024 · similarity 0.87)
- Divide and Conquer: Hybrid Pre-training for Person Search (AAAI 2024 · 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.
