Reproduce “FlexiReID: Adaptive Mixture of Expert for Multi-Modal 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
Zhen Sun, Lei Tan, Yunhang Shen, Chengmao Cai, Xing Sun, Pingyang Dai, Liujuan Cao, Rongrong Ji
FlexiReID introduces a unified framework for person re-identification capable of flexible retrieval across any combination of RGB, infrared, sketch, and text modalities. The system achieves state-of-the-art or competitive performance on various multimodal tasks, demonstrating enhanced robustness and accuracy from leveraging diverse information sources.
Key insights
- Flexible retrieval across arbitrary combinations of modalities, including handling missing inputs, significantly enhances person re-identification performance and real-world applicability.
- Dynamically allocating experts in a Mixture-of-Experts setup, rather than using a fixed number, optimizes feature extraction efficiency and effectiveness for heterogeneous multimodal data.
- The construction of a unified multimodal dataset (CIRS-PEDES) is crucial for comprehensive evaluation and comparison of flexible cross-modal Re-ID systems.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.