Person Re-Identification & Face Recognition
Robust Cross-Modality and Cross-Platform Retrieval for Person and Vehicle Re-Identification
Generated automatically from the limitations stated in 3 papers (NeurIPS, ICCV), listed under Evidence. It is not a paper, and it does not come from papers submitted to CSPaper.
The problem
Current cross-modality re-identification methods report relative improvements on benchmark datasets, yet their absolute retrieval performance drops severely in strongly mismatched sensor and platform scenarios (e.g., 9–16 mAP on RGBNT201 and MSVR310). When spectral shifts (RGB to Near-Infrared or Thermal) coincide with cross-platform viewpoint or domain shifts, identity discrimination fails at a fundamental level. This severe degradation blocks the practical deployment of automated re-identification systems across heterogeneous, multi-sensor surveillance networks operating in varying lighting and platform conditions.
Why it matters
Reliable automated person and vehicle re-identification across heterogeneous multi-sensor systems without requiring site-specific, cross-spectral manual annotation.
Ways to approach it
Prior-work checks are free with an account. Results someone already ran are shown to everyone.
- 1
Multi-Benchmark Robustness Evaluation: Systematically evaluate leading cross-modal and cross-spectral re-ID methods across cross-domain and multi-spectral splits (RGBNT201, MSVR310, RegDB, SYSU-MM01), measuring absolute mAP and Rank-1 drop when transferring models across mismatched sensor platforms without target fine-tuning.
- 2
Modality-Agnostic Identity Feature Disentanglement: Implement and test explicit disentanglement strategies that separate sensor-specific modality styles from identity-consistent biometric cues, measuring cross-spectral absolute mAP improvements on R-to-N and RT-to-N retrieval protocols.
- 3
Cross-Spectral Synthetic Augmentation: Adapt generative cross-modal synthesis pipelines to generate paired multi-spectral variations during training, measuring the resulting zero-shot transfer performance across disjoint camera networks and modalities.
Have a different approach?
Describe how you would tackle this problem and we'll look for papers that already do it. Free; your text stays private.
Why it might fail
If extreme cross-spectral shifts (such as thermal or nighttime NIR at low resolution) physically obscure distinguishing identity traits, absolute performance may be constrained by an insurmountable information-theoretic limit rather than representational deficiencies.
Evidence
Each paper's own statement of the limitation, verbatim.
- Unbiased Prototype Consistency Learning for Multi-Modal and Multi-Task Object Re-IdentificationNeurIPS 2025
Cross-domain absolute performance is very low (e.g., 9-14 mAP on MSVR310 cross-modal settings), indicating limited real-world generalization despite relative gains
- Multi-modal Multi-platform Person Re-Identification: Benchmark and MethodICCV 2025
Cross-modality & platform absolute performance remains low (~43 Rank-1, ~40 mAP), so the problem is far from solved
- MDReID: Modality-Decoupled Learning for Any-to-Any Multi-Modal Object Re-IdentificationNeurIPS 2025
Absolute performance in strongly mismatched scenarios remains low (e.g., 16.6 mAP R-to-N and 16.7 mAP RT-to-N on RGBNT201, 22.1 average mAP on MSVR310), indicating large cross-modality gaps persist
Nearest existing work
- Multi-modal Multi-platform Person Re-Identification: Benchmark and MethodICCV 2025
- RGB-Infrared Cross-Modality Person Re-IdentificationICCV 2017
- FUSE: Frequency-domain Unification and Spectral Energy Alignment for Multi-modal Object Re-IdentificationICML 2026
- Visible-Infrared Person Re-Indentification via Feature Fusion and Deep Mutual LearningACML 2024
- Cross-Modality Perturbation Synergy Attack for Person Re-identificationNeurIPS 2024
- RLE: A Unified Perspective of Data Augmentation for Cross-Spectral Re-IdentificationNeurIPS 2024
- RGB-Infrared Cross-Modality Person Re-Identification via Joint Pixel and Feature AlignmentICCV 2019
- Domain Shifting: A Generalized Solution for Heterogeneous Cross-Modality Person Re-IdentificationECCV 2024
- Modality Unifying Network for Visible-Infrared Person Re-IdentificationICCV 2023
- Multi-Memory Matching for Unsupervised Visible-Infrared Person Re-IdentificationECCV 2024
- VRCLIP: Multimodal Canonical Correlation Alignment for CLIP-Driven Vision-Radio Person Re-IdentificationCVPR 2026
- Cross-Modality Person Re-Identification via Modality Confusion and Center AggregationICCV 2021
- Learning to Reduce Dual-Level Discrepancy for Infrared-Visible Person Re-IdentificationCVPR 2019
- Learning Modal-Invariant and Temporal-Memory for Video-Based Visible-Infrared Person Re-IdentificationCVPR 2022
- Weakly Supervised Visible-Infrared Person Re-Identification via Heterogeneous Expert Collaborative Consistency LearningICCV 2025
Related open problems
Person Re-Identification & Face Recognition
Cross-Distribution Robustness and Granularity Evaluation in Cloth-Changing Person Re-Identification
Existing cloth-changing person re-identification (CC-ReID) and gait recognition models are trained and tested on datasets with restricted clothing variation profiles, such as upper-body-only changes or invariant footwear. Models trained on these biased distributions experience severe performance drops (such as single-digit Rank-1 accuracy) when evaluated on full clothing or pants-only changes. Without a standardized cross-setting evaluation that separates upper, lower, full, and footwear changes, reported benchmark metrics obscure localized overfitting and fail to measure true clothing-invariant representations.
Person Re-Identification & Face Recognition
Cross-Encoder and Target-Free Robustness Benchmarking for Facial Representations
Current facial representation and transfer methods rely on training-time preconditions that frequently fail in real-world deployment. Specifically, techniques depend on collecting unlabeled target-domain demographic data—which is often legally or ethically prohibited—or evaluate strictly against known, seen feature extractors within face-only datasets. When the deployed encoder is changed or target data cannot be sampled in advance, systems experience substantial performance drops that remain unquantified across modern architectures. Without a unified benchmark testing methods under zero-target-data and unseen-encoder constraints, practitioners cannot determine which representation strategies legitimately transfer.