DupNet: Towards Very Tiny Quantized CNN with Improved Accuracy for Face Detection

TL;DR

This paper introduces DupNet, a highly compact quantized CNN for face detection on edge devices, which significantly reduces model size and complexity while maintaining high accuracy by duplicating channels and feature maps.

Contribution

The paper proposes DupNet, a novel approach combining weight and feature map duplication to create the smallest deep face detector with improved accuracy for edge deployment.

Findings

DupNet-Tinier-YOLO is 6.5X smaller than IFQ-Tinier-YOLO.

DupNet-Tinier-YOLO achieves 2.4% higher detection accuracy.

Our model is only 36.9 KB, the smallest deep face detector to date.

Abstract

Deploying deep learning based face detectors on edge devices is a challenging task due to the limited computation resources. Even though binarizing the weights of a very tiny network gives impressive compactness on model size (e.g. 240.9 KB for IFQ-Tinier-YOLO), it is not tiny enough to fit in the embedded devices with strict memory constraints. In this paper, we propose DupNet which consists of two parts. Firstly, we employ weights with duplicated channels for the weight-intensive layers to reduce the model size. Secondly, for the quantization-sensitive layers whose quantization causes notable accuracy drop, we duplicate its input feature maps. It allows us to use more weights channels for convolving more representative outputs. Based on that, we propose a very tiny face detector, DupNet-Tinier-YOLO, which is 6.5X times smaller on model size and 42.0% less complex on computation and meanwhile achieves 2.4% higher detection than IFQ-Tinier-YOLO. Comparing with the full precision Tiny-YOLO, our DupNet-Tinier-YOLO gives 1,694.2X and 389.9X times savings on model size and computation complexity respectively with only 4.0% drop on detection rate (0.880 vs. 0.920). Moreover, our DupNet-Tinier-YOLO is only 36.9 KB, which is the tiniest deep face detector to our best knowledge.