Masked Face Inpainting Through Residual Attention UNet

TL;DR

This paper introduces a residual attention UNet model for blind face mask inpainting that effectively restores facial details by addressing gradient vanishing and focusing on mask regions, outperforming existing methods.

Contribution

The paper proposes a novel residual attention UNet architecture for mask face inpainting that improves detail restoration and training stability without external masks.

Findings

Model achieves high-fidelity face restoration on CelebA dataset.

Residual blocks mitigate vanishing gradient issues.

Attention units improve focus on mask regions, reducing resource use.

Abstract

Realistic image restoration with high texture areas such as removing face masks is challenging. The state-of-the-art deep learning-based methods fail to guarantee high-fidelity, cause training instability due to vanishing gradient problems (e.g., weights are updated slightly in initial layers) and spatial information loss. They also depend on intermediary stage such as segmentation meaning require external mask. This paper proposes a blind mask face inpainting method using residual attention UNet to remove the face mask and restore the face with fine details while minimizing the gap with the ground truth face structure. A residual block feeds info to the next layer and directly into the layers about two hops away to solve the gradient vanishing problem. Besides, the attention unit helps the model focus on the relevant mask region, reducing resources and making the model faster. Extensive experiments on the publicly available CelebA dataset show the feasibility and robustness of our proposed model. Code is available at \url{https://github.com/mdhosen/Mask-Face-Inpainting-Using-Residual-Attention-Unet}