Perspective-assisted prototype-based learning for semi-supervised crowd counting

Yifei Qian, Liangfei Zhang, Zhongliang Guo, Xiaopeng Hong, Ognjen Arandjelovic, Carl R. Donovan

Abstract: Highlights•A novel perspective-assisted prototype-based learning approach for semi-supervised crowd counting.•A novel perspective pseudo-label assigner based on regional perspective self-organization, to provide perspective classes for image patches.•Extensive experiments on four crowd benchmarks to validate the effectiveness of our semi-supervised approach.