Reproduce “X-ReID: Multi-granularity Information Interaction for Video-Based 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
Chenyang Yu, Xuehu Liu, Pingping Zhang, Huchuan Lu
The X-ReID framework enhances video-based visible-infrared person re-identification by employing Cross-modality Prototype Collaboration for modality gap reduction and Multi-granularity Information Interaction for robust spatiotemporal feature learning. It achieves top performance on HITSZ-VCM with 73.4% Rank-1 (I2V) and BUPTCampus with 68.2% Rank-1 (I2V), surpassing existing methods.
Key insights
- Cross-modality prototype collaboration effectively transfers CLIP's pre-trained knowledge and narrows the modality gap in a text-free manner by fusing information from opposite modalities into prototypes.
- Multi-granularity information interaction, encompassing both short-term (adjacent frames) and long-term (wider temporal strides) dependencies, significantly enhances the robustness of spatiotemporal feature learning.
- Explicitly integrating cross-modality alignment directly within the temporal modeling pipeline, rather than as a separate step, produces more discriminative features across visible and infrared modalities.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.