Learning Deep Features via Congenerous Cosine Loss for Person Recognition
This addresses person recognition in complex scenes for applications like surveillance, but it appears incremental as it builds on existing feature learning approaches.
The paper tackles person recognition by training a network to produce robust features using a congenerous cosine loss that minimizes cosine distance between samples and their cluster centroids, achieving better classification accuracy compared to previous state-of-the-art methods.
Person recognition aims at recognizing the same identity across time and space with complicated scenes and similar appearance. In this paper, we propose a novel method to address this task by training a network to obtain robust and representative features. The intuition is that we directly compare and optimize the cosine distance between two features - enlarging inter-class distinction as well as alleviating inner-class variance. We propose a congenerous cosine loss by minimizing the cosine distance between samples and their cluster centroid in a cooperative way. Such a design reduces the complexity and could be implemented via softmax with normalized inputs. Our method also differs from previous work in person recognition that we do not conduct a second training on the test subset. The identity of a person is determined by measuring the similarity from several body regions in the reference set. Experimental results show that the proposed approach achieves better classification accuracy against previous state-of-the-arts.