Investigation of Architectures and Receptive Fields for Appearance-based Gaze Estimation

TL;DR

This paper demonstrates that simple modifications to a ResNet architecture, such as adjusting stride, resolution, and multi-region input, can outperform complex existing methods in appearance-based gaze estimation across multiple datasets.

Contribution

The study reveals that tuning basic ResNet parameters significantly improves gaze estimation accuracy, highlighting the importance of architecture choices over complex mechanisms.

Findings

ResNet-50 achieves state-of-the-art results on three datasets.

Stride, resolution, and multi-region architecture are critical for performance.

Simple architecture tuning outperforms many complex methods.

Abstract

With the rapid development of deep learning technology in the past decade, appearance-based gaze estimation has attracted great attention from both computer vision and human-computer interaction research communities. Fascinating methods were proposed with variant mechanisms including soft attention, hard attention, two-eye asymmetry, feature disentanglement, rotation consistency, and contrastive learning. Most of these methods take the single-face or multi-region as input, yet the basic architecture of gaze estimation has not been fully explored. In this paper, we reveal the fact that tuning a few simple parameters of a ResNet architecture can outperform most of the existing state-of-the-art methods for the gaze estimation task on three popular datasets. With our extensive experiments, we conclude that the stride number, input image resolution, and multi-region architecture are critical for the gaze estimation performance while their effectiveness dependent on the quality of the input face image. We obtain the state-of-the-art performances on three datasets with 3.64 on ETH-XGaze, 4.50 on MPIIFaceGaze, and 9.13 on Gaze360 degrees gaze estimation error by taking ResNet-50 as the backbone.