Research map / Label-efficient and robust learning / Recognition and long-tail
Shape-Erased Feature Learning for Visible-Infrared Person Re-Identification
Jiawei Feng, Ancong Wu, Wei-Shi Zheng
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 |
|---|---|---|---|
| Use shape-related objective without shape-erased objective (ablation) | SGIEL · SYSU-MM01 all search, single-shot | Rank-1: 70.73 -> 69.60 | Hurt |
| Add shape-erased objective alongside shape-related objective (ablation) | SGIEL · SYSU-MM01 all search, single-shot | Rank-1: 69.60 -> 72.00 | Improved |
| Add orthogonality loss (ablation) | SGIEL with shape-related and shape-erased objectives · SYSU-MM01 all search, single-shot | Rank-1: 72.00 -> 73.52 | Improved |
| Add gradient-based reweighting of shape objectives (ablation) | SGIEL · SYSU-MM01 all search, single-shot | Rank-1: 73.52 -> 75.18 | Improved |
| Replace one orthogonal projector with two unconstrained projectors (ablation) | SGIEL · SYSU-MM01 all search, single-shot | Rank-1: 75.18 -> 70.18 | Hurt |
| Remove orthogonality loss while keeping decomposition (ablation) | SGIEL · SYSU-MM01 all search, single-shot | Rank-1: 75.18 -> 73.49 | Hurt |
| Add orthogonality loss between two independent projectors (ablation) | Two-projector variant · SYSU-MM01 all search, single-shot | Rank-1: 70.18 -> 72.76 | Improved |
| Use one projector rather than two independent projectors, without orthogonality loss (ablation) | Shape-erased feature learning · SYSU-MM01 all search, single-shot | Rank-1: 70.18 -> 73.49 | Improved |
| Concatenate outputs of two checkpoints | Ours · SYSU-MM01 all search, single-shot | Rank-1: 75.18 -> 77.12 | Improved |
| Use SGIEL rather than the classification-and-triplet baseline | HITSZ-VCM · HITSZ-VCM infrared to visible | Rank-1: 62.02 -> 67.65 | Improved |
| Use SGIEL rather than the classification-and-triplet baseline | HITSZ-VCM · HITSZ-VCM visible to infrared | Rank-1: 64.90 -> 70.23 | Improved |
Closest papers by meaning
- Modality Unifying Network for Visible-Infrared Person Re-Identification (ICCV 2023 · similarity 0.91)
- Learning Progressive Modality-Shared Transformers for Effective Visible-Infrared Person Re-identification (AAAI 2023 · similarity 0.90)
- Implicit Discriminative Knowledge Learning for Visible-Infrared Person Re-Identification (CVPR 2024 · similarity 0.90)
- High-Order Structure Based Middle-Feature Learning for Visible-Infrared Person Re-identification (AAAI 2024 · similarity 0.89)
- Empowering Visible-Infrared Person Re-Identification with Large Foundation Models (NeurIPS 2024 · similarity 0.89)
- Shallow-Deep Collaborative Learning for Unsupervised Visible-Infrared Person Re-Identification (CVPR 2024 · similarity 0.88)
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.
