Research map / Label-efficient and robust learning / Recognition and long-tail
Residual Compensation Networks for Heterogeneous Face Recognition
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 MD loss to RC module (ablation) | RCN · IIIT-D Viewed Sketch | Rank-1: 88.24 -> 90.34 | Helped |
| add MD loss to RC module (ablation) | RCN · CASIA NIR-VIS 2.0 | Rank-1: 98.91 ± 0.17 -> 99.32 ± 0.15 | Helped |
| replace MD loss with center loss (ablation) | RCN · IIIT-D Viewed Sketch | Rank-1: 90.34 -> 86.55 | Hurt |
| replace MD loss with contrastive loss (ablation) | RCN · IIIT-D Viewed Sketch | Rank-1: 90.34 -> 88.66 | Hurt |
| add PReLU nonlinearity to RC module (ablation) | RCN · IIIT-D Viewed Sketch | Rank-1: 88.24 -> 90.34 | Helped |
| add PReLU nonlinearity to RC module (ablation) | RCN · CASIA NIR-VIS 2.0 | Rank-1: 98.82 ± 0.35 -> 99.32 ± 0.15 | Helped |
| stack two RC modules instead of one (ablation) | RCN · IIIT-D Viewed Sketch | Rank-1: 90.34 -> 89.50 | Hurt |
| add RC module to all layers (ablation) | RCN · CASIA NIR-VIS 2.0 | Rank-1: 99.32 -> 98.81 | Hurt |
| use residual compensation instead of plain extra FC branch (c) (ablation) | RCN · IIIT-D Viewed Sketch | Rank-1: 76.47 -> 88.24 | Helped |
| freeze backbone and fine-tune only FC/RC versus fine-tune all layers (ablation) | ResNet-10 · IIIT-D Viewed Sketch | Rank-1: 82.35 -> 86.97 | Helped |
Closest papers by meaning
- Dual Variational Generation for Low Shot Heterogeneous Face Recognition (NeurIPS 2019 · similarity 0.86)
- Adversarial Discriminative Heterogeneous Face Recognition (AAAI 2018 · similarity 0.86)
- Asymmetric Joint Learning for Heterogeneous Face Recognition (AAAI 2018 · similarity 0.86)
- UMFN: Unified Multi-Domain Face Normalization for Joint Cross-domain Prototype Learning and Heterogeneous Face Recognition (CVPR 2025 · similarity 0.85)
- Heterogeneous Test-Time Training for Multi-Modal Person Re-identification (AAAI 2024 · similarity 0.82)
- Multi-Modal Object Re-identification via Sparse Mixture-of-Experts (ICML 2025 · similarity 0.82)
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.