Compressing Facial Makeup Transfer Networks by Collaborative Distillation and Kernel Decomposition

TL;DR

This paper proposes a novel approach to compress facial makeup transfer networks using collaborative distillation and kernel decomposition, significantly reducing model size while maintaining high-quality image generation.

Contribution

It introduces a new collaborative distillation method based on encoder-decoder relationships and applies kernel decomposition to create a lightweight CNN for makeup transfer.

Findings

Effective compression of BeautyGAN with minimal quality loss

Significant reduction in model size and computational cost

Maintains perceptual quality of generated makeup images

Abstract

Although the facial makeup transfer network has achieved high-quality performance in generating perceptually pleasing makeup images, its capability is still restricted by the massive computation and storage of the network architecture. We address this issue by compressing facial makeup transfer networks with collaborative distillation and kernel decomposition. The main idea of collaborative distillation is underpinned by a finding that the encoder-decoder pairs construct an exclusive collaborative relationship, which is regarded as a new kind of knowledge for low-level vision tasks. For kernel decomposition, we apply the depth-wise separation of convolutional kernels to build a light-weighted Convolutional Neural Network (CNN) from the original network. Extensive experiments show the effectiveness of the compression method when applied to the state-of-the-art facial makeup transfer network -- BeautyGAN.