Aug 2026· International journal of pattern recognition and artificial intelligence· 0 citations
TL;DR
A Progressively Biased Split Vision Transformer (PBSVT) is proposed, which combines a split ViT backbone with progressive bias training to gradually reduce RGB-dominant bias while preserving modality-shared structure and demonstrates the effectiveness of progressive modality transition for robust VI-ReID representation learning.
Abstract
Visible-Infrared Person Re-Identification (VI-ReID) remains a challenging task due to the significant modality gap between visible and infrared images, which hinders accurate cross-modality identity matching. Existing methods often struggle to balance modality invariance and feature discriminability. Most methods employ ImageNet-pretrained backbones that are heavily biased toward RGB statistics, causing the extracted cross-modal features to over-emphasize visible-spectrum information and weakening the understanding of infrared cues. To address this issue, some works introduce third-modality generation or grayscale-based augmentation, but these strategies either increase training complexity or still leave a non-negligible discrepancy from real infrared data. We propose a Progressively Biased Split Vision Transformer (PBSVT), which combines a split ViT backbone with progressive bias training to gradually reduce RGB-dominant bias while preserving modality-shared structure. Extensive experiments on SYSU-MM01 and RegDB show that PBSVT achieves state-of-the-art or highly competitive performance while introducing no additional inference cost. PBSVT obtains the best results on 8 of the 12 reported indicators, including 77.90% Rank-1 and 97.98% Rank-10 on SYSU-MM01 all-search, 83.74% Rank-1 on SYSU-MM01 indoor-search, and 92.89% Rank-1, 98.77% Rank-10, and 92.18% mAP on RegDB Visible-to-Infrared. These results demonstrate the effectiveness of progressive modality transition for robust VI-ReID representation learning.
This work proposes MDCRNet, a Multi-scale Decomposed Convolution Refinement Network that enhances cross-modal feature learning and discriminative metric learning, and develops a Joint Discriminative Metric Loss incorporating a novel Granularity Discriminative Loss (GDL).
Mingsheng Zheng, Zirui Jiang, Bo Liu et al.· 0 citations
Visible-infrared person re-identification (VI-ReID) remains challenging due to severe spectral discrepancy and local cross-modal misalignment between visible and infrared images. Most existing methods alleviate this discrepancy through middle- or late-stage feature alignment, but modality-specific shallow features may have already accumulated spectral bias and local correspondence errors before shared representations are formed. In this paper, we argue that VI-ReID should be treated as an early cross-modal correspondence learning problem rather than only a late embedding alignment problem. To this end, we propose \textbf{CMIA-Net}, a framework that establishes bidirectional visible-infrared interaction at shallow backbone stages. Its core module, \textbf{Cross-Modal Interaction Attention (CMIA)}, enables visible and infrared feature maps to exchange complementary local information before deep semantic aggregation, thereby reducing progressive stream divergence. To stabilize early interaction, we further introduce \textbf{Spectral-Invariant Augmentation (MC-Aug)} to suppress over-reliance on visible-spectrum cues and a \textbf{Bi-directional Hetero-Center Learning (BHCT)} loss to improve class-center-level cross-modal compactness and inter-class separability. Experiments on RegDB and SYSU-MM01 show that CMIA-Net achieves strong performance on RegDB and the SYSU-MM01 indoor-search protocol, while remaining competitive under the more challenging SYSU-MM01 all-search setting. Specifically, CMIA-Net obtains 94.09\% Rank-1 accuracy and 90.74\% mAP in the RegDB visible-to-thermal setting, and 66.67\% Rank-1 accuracy and 58.12\% mAP in the SYSU-MM01 all-search setting.
Unknown authors· Engineering Research Express· 0 citations
Infrared and visible image fusion is pivotal for robust visual perception across all weather conditions and scenes. Although deep learning-based methods have made notable progress, most either assume pre-aligned inputs or rely on implicit feature-space alignment, which fails to fundamentally address the amplification of registration errors and the loss of semantic structure in the fused results. To this end, we propose a universal representation and end-to-end framework for jointly registering and fusing unaligned infrared-visible image pairs, dubbed URMIF. Each image is mapped into modality-invariant (homogeneous) and modality-specific (heterogeneous) features: the invariant "structural skeleton" encodes geometry and semantics to stabilize alignment, while the specific "texture carrier" preserves thermal saliency and visible details to enable complementary fusion. Therefore, we propose a bi-directionally coupled registration-fusion module. This module performs hierarchical deformation estimation from coarse to fine, effectively mitigating visual mismatches caused by complex parallax in real-world scenes. Within this framework, the fusion component acts as the "evaluator" of registration, providing feedback regularization to update the deformation and suppress error accumulation. Furthermore, we introduce a dominant-plane prior as a scene-level constraint, seeding stable global and patch-wise homographies and reconciling cross-modal detail conflicts, to reinforce geometric consistency and semantic reliability. We also release a large-scale dataset comprising 1,500+ unaligned infrared/visible pairs with registration ground truth, spanning diverse illumination conditions and fields of view. Based on this dataset and additional benchmarks, extensive experiments validate that our framework achieves robust alignment and high-quality fusion on misaligned inputs, markedly reducing artifacts and improving the performance of downstream tasks such as detection and segmentation. Code and benchmark are available at https://github.com/ZengxiZhang/URMIF.
Jinyuan Liu, Zengxi Zhang, Jiahao Zhang et al.· IEEE Transactions on Pattern...· 0 citations
A Transformer-based baseline framework for visible-infrared ReID is proposed, designed to effectively capture modality-invariant features and outline several promising directions for future research.
Xiao Wang, Bing Wang, Bin Yang et al.· arXiv.org· 0 citations
Unsupervised person re-identification (USL-ReID) typically relies on clustering to generate pseudo-labels, but significant cross-view appearance variations often cause images of the same identity to be split into different clusters. Training on such noisy pseudo-labels severely degrades the learned representations. Therefore, learning robust view-invariant features is paramount. Data augmentation provides a direct way to enhance invariance, yet its trade-offs in USL-ReID remain under-explored: weak augmentations usually preserve identity semantics but lack diversity, whereas strong augmentations provide richer appearance diversity at the cost of partially corrupting identity-consistent semantic cues. To address this challenge, we propose Invariant Representation learning with Progressive Prototype Refinement (IRPP), a unified framework that learns invariant and discriminative features from noisy pseudo-labels. IRPP consists of three synergistic components. First, an Augmented Dual-Contrastive Learning (ADCL) module performs dataset-level prototype-guided invariant learning by contrasting weakly and strongly augmented views against cluster-derived prototypes. Second, an Alignment and Uniformity Learning (AUL) module regularizes the mini-batch-level weak–strong feature geometry, leading to more stable feature distributions under data augmentation. Third, a Progressive Prototype Refinement (PPR) mechanism progressively optimizes cluster centroids into cleaner prototypes, thereby mitigating the influence of noisy pseudo-labels and further strengthening invariant representation learning. This closed-loop design enables prototype-guided contrastive learning, weak–strong regularization, and prototype refinement to mutually reinforce each other. Extensive experiments on standard USL-ReID benchmarks demonstrate that IRPP achieves state-of-the-art performance with a simple and efficient training pipeline. Code is available at https://github.com/Trangle12/IRPP
Xuan Tan, Qixian Zhang, Ding Qi et al.· IEEE Transactions on Image P...· 0 citations
Person re-identification (Re-ID) aims to retrieve the same person across non-overlapping cameras. Despite recent progress, Re-ID remains highly challenging due to high inter-class similarity in large-scale datasets and drastic intra-class variations in cross-platform scenarios (e.g., drones and wearable cameras). While Transformers have been introduced to Re-ID for their superior long-range dependency modeling and robustness against global appearance morphology changes, they often struggle to capture fine-grained discriminative local features due to the inherent properties of the self-attention mechanism. To address this, we propose a Reinforced-Transformer (RT) architecture designed to recover and reinforce these critical local details. Specifically, we develop a Detail Retention Module (DRM) to preserve salient local information while enhancing the interaction between local and global features. Building on this, a Multi-Detail Recovery Module (MDRM) is introduced to progressively restore local features from both coarse-to-fine and fine-to-coarse perspectives. Extensive experiments on four large-scale benchmarks and a mixed aerial-ground dataset demonstrate that our method achieves state-of-the-art (SOTA) or highly competitive performance across different benchmarks.
Meifeng Liu, Hua Han, A. A. M. Muzahid et al.· IEEE Access· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
