An Effective Deep Network for Head Pose Estimation without Keypoints

TL;DR

This paper introduces a lightweight, real-time deep learning model for head pose estimation that leverages knowledge distillation from teacher models trained on synthetic data, achieving high accuracy and efficiency.

Contribution

The paper presents a novel, compact head pose estimation model using ResNet18 and knowledge distillation from synthetic data-trained teachers, improving accuracy and speed.

Findings

Significantly outperforms existing methods in accuracy.

Operates at approximately 300 FPS on Tesla V100.

Effective on real-world datasets AFLW-2000 and BIWI.

Abstract

Human head pose estimation is an essential problem in facial analysis in recent years that has a lot of computer vision applications such as gaze estimation, virtual reality, and driver assistance. Because of the importance of the head pose estimation problem, it is necessary to design a compact model to resolve this task in order to reduce the computational cost when deploying on facial analysis-based applications such as large camera surveillance systems, AI cameras while maintaining accuracy. In this work, we propose a lightweight model that effectively addresses the head pose estimation problem. Our approach has two main steps. 1) We first train many teacher models on the synthesis dataset - 300W-LPA to get the head pose pseudo labels. 2) We design an architecture with the ResNet18 backbone and train our proposed model with the ensemble of these pseudo labels via the knowledge distillation process. To evaluate the effectiveness of our model, we use AFLW-2000 and BIWI - two real-world head pose datasets. Experimental results show that our proposed model significantly improves the accuracy in comparison with the state-of-the-art head pose estimation methods. Furthermore, our model has the real-time speed of $\sim$300 FPS when inferring on Tesla V100.