Find Gold in Sand: Fine-Grained Similarity Mining for Domain-Adaptive Crowd Counting

Abstract

The domain shift of crowd scenes significantly hinders the application of crowd counting models in open scenarios. Although domain adaptation methods for crowd counting have bridged this gap to some extent, they ignore one of the significant causes of domain shift, which is the inter-domain data distribution bias. We discover a connection between known and unknown distributions, offering an opportunity for similarity mining to address domain shift. However, there are still challenges related to insufficient and inaccurate similarity mining. In this paper, we propose a novel Fine-grained Inter-domain Similarity Mining (FSIM) framework. To comprehensively explore similar distributions between source and target domains, we propose a Multi-scale Distribution Alignment (MDA) module based on diffusion retrieval. To enhance the reliability of inter-domain similarity mining, we propose a Multi-retrieval Refinement (MR) module based on evidence theory, serving as an uncertainty measurement method. Eventually, to eliminate data distribution bias, we perform model retraining using similar distributions. Extensive experiments conducted on five standard crowd counting benchmarks, SHA, SHB, QNRF, NWPU, and JHU-CROWD++, demonstrate that the proposed FSIM has robust generalizability. Our code will be available at: https://github.com/HopooLinZ/FSIM/