Facial Landmark Points Detection Using Knowledge Distillation-Based Neural Networks

TL;DR

This paper introduces a knowledge distillation approach with a novel loss function to train lightweight neural networks for facial landmark detection, achieving high accuracy suitable for mobile devices.

Contribution

It proposes a new loss function and a dual-teacher knowledge distillation framework to improve lightweight facial landmark detection models.

Findings

Enhanced accuracy of lightweight models on facial landmark datasets

Effective use of dual teachers for guiding student networks

Achieved competitive results on challenging datasets

Abstract

Facial landmark detection is a vital step for numerous facial image analysis applications. Although some deep learning-based methods have achieved good performances in this task, they are often not suitable for running on mobile devices. Such methods rely on networks with many parameters, which makes the training and inference time-consuming. Training lightweight neural networks such as MobileNets are often challenging, and the models might have low accuracy. Inspired by knowledge distillation (KD), this paper presents a novel loss function to train a lightweight Student network (e.g., MobileNetV2) for facial landmark detection. We use two Teacher networks, a Tolerant-Teacher and a Tough-Teacher in conjunction with the Student network. The Tolerant-Teacher is trained using Soft-landmarks created by active shape models, while the Tough-Teacher is trained using the ground truth (aka Hard-landmarks) landmark points. To utilize the facial landmark points predicted by the Teacher networks, we define an Assistive Loss (ALoss) for each Teacher network. Moreover, we define a loss function called KD-Loss that utilizes the facial landmark points predicted by the two pre-trained Teacher networks (EfficientNet-b3) to guide the lightweight Student network towards predicting the Hard-landmarks. Our experimental results on three challenging facial datasets show that the proposed architecture will result in a better-trained Student network that can extract facial landmark points with high accuracy.