STNet: Scale Tree Network with Multi-level Auxiliator for Crowd Counting

TL;DR

STNet introduces a hierarchical scale tree structure and multi-level auxiliary learning to improve crowd counting accuracy amidst scale variation and complex backgrounds, achieving superior results on multiple datasets.

Contribution

The paper proposes a novel Scale Tree Network with a diversity enhancer and multi-level auxiliator for improved crowd counting accuracy, addressing scale variation and background complexity.

Findings

Outperforms existing methods on four challenging datasets.

Effectively handles scale variation and density inconsistency.

End-to-end training without manual loss weight tuning.

Abstract

Crowd counting remains a challenging task because the presence of drastic scale variation, density inconsistency, and complex background can seriously degrade the counting accuracy. To battle the ingrained issue of accuracy degradation, we propose a novel and powerful network called Scale Tree Network (STNet) for accurate crowd counting. STNet consists of two key components: a Scale-Tree Diversity Enhancer and a Semi-supervised Multi-level Auxiliator. Specifically, the Diversity Enhancer is designed to enrich scale diversity, which alleviates limitations of existing methods caused by insufficient level of scales. A novel tree structure is adopted to hierarchically parse coarse-to-fine crowd regions. Furthermore, a simple yet effective Multi-level Auxiliator is presented to aid in exploiting generalisable shared characteristics at multiple levels, allowing more accurate pixel-wise background cognition. The overall STNet is trained in an end-to-end manner, without the needs for manually tuning loss weights between the main and the auxiliary tasks. Extensive experiments on four challenging crowd datasets demonstrate the superiority of the proposed method.