Progressive Multi-resolution Loss for Crowd Counting

TL;DR

This paper introduces a multi-resolution loss function for crowd counting that improves density map regression by measuring at multiple resolutions, leading to better performance across several datasets.

Contribution

It proposes a novel multi-resolution L2 loss function for crowd counting, mathematically proven to outperform traditional single-resolution loss.

Findings

Significant performance improvements on four crowd counting datasets.

The multi-resolution loss outperforms state-of-the-art methods.

The approach is simple yet effective across different models.

Abstract

Crowd counting is usually handled in a density map regression fashion, which is supervised via a L2 loss between the predicted density map and ground truth. To effectively regulate models, various improved L2 loss functions have been proposed to find a better correspondence between predicted density and annotation positions. In this paper, we propose to predict the density map at one resolution but measure the density map at multiple resolutions. By maximizing the posterior probability in such a setting, we obtain a log-formed multi-resolution L2-difference loss, where the traditional single-resolution L2 loss is its particular case. We mathematically prove it is superior to a single-resolution L2 loss. Without bells and whistles, the proposed loss substantially improves several baselines and performs favorably compared to state-of-the-art methods on four crowd counting datasets, ShanghaiTech A & B, UCF-QNRF, and JHU-Crowd++.