Reproduce “Multi-scale Decomposed Convolution Refinement Network for Visible-Infrared Person Re-Identification” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Mingsheng Zheng, Zirui Jiang, Bo Liu, Yupeng Chen, Jun Zhang, Kai Zhao
The Multi-scale Decomposed Convolution Refinement Network (MDCRNet) advances visible-infrared person re-identification by integrating a Hierarchical Learning Module for multi-scale feature extraction and a novel Joint Discriminative Metric Loss for comprehensive optimization. The framework achieved state-of-the-art performance, with a 1.06% mAP gain on the RegDB dataset and a 0.75% mAP gain on the SYSU-MM01 dataset compared to prior methods.
Key insights
- Capturing features across multiple spatial scales (fine-grained, mid-grained, coarse-grained) using progressive dilated convolutions significantly enhances identity representation for VI-ReID.
- A unified metric learning objective, simultaneously enforcing intra-identity compactness, inter-identity separability, and cross-modal alignment, leads to more robust and discriminative feature learning.
- Hierarchical fusion combining channel attention with multi-scale spatial processing effectively mitigates cross-modal discrepancies and extracts comprehensive semantic information.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.