DASECount: Domain-Agnostic Sample-Efficient Wireless Indoor Crowd Counting via Few-shot Learning

TL;DR

DASECount is a novel domain-agnostic, sample-efficient wireless indoor crowd counting framework that leverages few-shot learning and CNN-based feature extraction to achieve high cross-domain accuracy with minimal data.

Contribution

It introduces a two-stage few-shot learning approach with CNN modules and knowledge distillation for robust cross-domain indoor crowd counting.

Findings

Achieves over 92.68% accuracy in 0-8 people counting tasks.

Significantly outperforms benchmark methods in cross-domain scenarios.

Effective with as few as 5 target domain samples.

Abstract

Accurate indoor crowd counting (ICC) is a key enabler to many smart home/office applications. In this paper, we propose a Domain-Agnostic and Sample-Efficient wireless indoor crowd Counting (DASECount) framework that suffices to attain robust cross-domain detection accuracy given very limited data samples in new domains. DASECount leverages the wisdom of few-shot learning (FSL) paradigm consisting of two major stages: source domain meta training and target domain meta testing. Specifically, in the meta-training stage, we design and train two separate convolutional neural network (CNN) modules on the source domain dataset to fully capture the implicit amplitude and phase features of CSI measurements related to human activities. A subsequent knowledge distillation procedure is designed to iteratively update the CNN parameters for better generalization performance. In the meta-testing stage, we use the partial CNN modules to extract low-dimension features out of the high-dimension input target domain CSI data. With the obtained low-dimension CSI features, we can even use very few shots of target domain data samples (e.g., 5-shot samples) to train a lightweight logistic regression (LR) classifier, and attain very high cross-domain ICC accuracy. Experiment results show that the proposed DASECount method achieves over 92.68\%, and on average 96.37\% detection accuracy in a 0-8 people counting task under various domain setups, which significantly outperforms the other representative benchmark methods considered.