Crowd Counting Using Scale-Aware Attention Networks

In this paper, we consider the problem of crowd counting in images. Given an image of a crowded scene, our goal is to estimate the density map of this image, where each pixel value in the density map corresponds to the crowd density at the corresponding location in the image. Given the estimated density map, the final crowd count can be obtained by summing over all values in the density map. One challenge of crowd counting is the scale variation in images. In this work, we propose a novel scale-aware attention network to address this challenge. Using the attention mechanism popular in recent deep learning architectures, our model can automatically focus on certain global and local scales appropriate for the image. By combining these global and local scale attentions, our model outperforms other state-of-the-art methods for crowd counting on several benchmark datasets.

1. Introduction

We consider the problem of crowd counting in arbitrary static images. Given an arbitrary image of a crowded scene without any prior knowledge about the scene (e.g. camera position, scene layout, crowd density), our goal is to estimate the density map of the input image, where each pixel value in the density map corresponds to the crowd density at the corresponding location of the input image. The crowd count can be obtained by integrating the entire density map. In particular, we focus on the setting where the training data have dotted annotations, i.e. each object instance (e.g. people) is annotated with a single point in the image.