Active Crowd Counting with Limited Supervision

TL;DR

This paper introduces an active learning framework for crowd counting that reduces annotation effort by selecting the most informative images and leveraging unlabeled data, achieving near state-of-the-art performance with limited labels.

Contribution

The paper proposes a novel active learning approach with diverse sample selection and distribution alignment techniques for crowd counting with limited supervision.

Findings

Achieves near state-of-the-art accuracy with only 10% of labeled data.

Effectively utilizes unlabeled data through distribution alignment and mix-up strategies.

Demonstrates strong performance on multiple benchmark datasets.

Abstract

To learn a reliable people counter from crowd images, head center annotations are normally required. Annotating head centers is however a laborious and tedious process in dense crowds. In this paper, we present an active learning framework which enables accurate crowd counting with limited supervision: given a small labeling budget, instead of randomly selecting images to annotate, we first introduce an active labeling strategy to annotate the most informative images in the dataset and learn the counting model upon them. The process is repeated such that in every cycle we select the samples that are diverse in crowd density and dissimilar to previous selections. In the last cycle when the labeling budget is met, the large amount of unlabeled data are also utilized: a distribution classifier is introduced to align the labeled data with unlabeled data; furthermore, we propose to mix up the distribution labels and latent representations of data in the network to particularly improve the distribution alignment in-between training samples. We follow the popular density estimation pipeline for crowd counting. Extensive experiments are conducted on standard benchmarks i.e. ShanghaiTech, UCF CC 50, MAll, TRANCOS, and DCC. By annotating limited number of images (e.g. 10% of the dataset), our method reaches levels of performance not far from the state of the art which utilize full annotations of the dataset.