A Human Ear Reconstruction Autoencoder

TL;DR

This paper introduces HERA, a self-supervised autoencoder framework for 3D human ear reconstruction from 2D images, addressing the challenge of subtle ear features and in-the-wild image variability.

Contribution

It presents the first self-supervised autoencoder model for 3D ear reconstruction, incorporating an in-the-wild ear color model to handle diverse image conditions.

Findings

Effective 3D ear pose and shape prediction without supervision

Good performance on 2D landmark localization

Realistic 3D ear reconstructions from in-the-wild images

Abstract

The ear, as an important part of the human head, has received much less attention compared to the human face in the area of computer vision. Inspired by previous work on monocular 3D face reconstruction using an autoencoder structure to achieve self-supervised learning, we aim to utilise such a framework to tackle the 3D ear reconstruction task, where more subtle and difficult curves and features are present on the 2D ear input images. Our Human Ear Reconstruction Autoencoder (HERA) system predicts 3D ear poses and shape parameters for 3D ear meshes, without any supervision to these parameters. To make our approach cover the variance for in-the-wild images, even grayscale images, we propose an in-the-wild ear colour model. The constructed end-to-end self-supervised model is then evaluated both with 2D landmark localisation performance and the appearance of the reconstructed 3D ears.