PoNA: Pose-guided Non-local Attention for Human Pose Transfer

TL;DR

This paper introduces PoNA, a pose-guided non-local attention mechanism within a GAN framework, to improve human pose transfer by capturing long-range dependencies, resulting in sharper, more realistic images with fewer parameters.

Contribution

The paper proposes a novel pose-guided non-local attention mechanism and a simplified cascaded GAN architecture for improved human pose transfer.

Findings

Produces sharper, more realistic images with rich details.

Achieves faster speed and fewer parameters than previous methods.

Helps alleviate data insufficiency in person re-identification.

Abstract

Human pose transfer, which aims at transferring the appearance of a given person to a target pose, is very challenging and important in many applications. Previous work ignores the guidance of pose features or only uses local attention mechanism, leading to implausible and blurry results. We propose a new human pose transfer method using a generative adversarial network (GAN) with simplified cascaded blocks. In each block, we propose a pose-guided non-local attention (PoNA) mechanism with a long-range dependency scheme to select more important regions of image features to transfer. We also design pre-posed image-guided pose feature update and post-posed pose-guided image feature update to better utilize the pose and image features. Our network is simple, stable, and easy to train. Quantitative and qualitative results on Market-1501 and DeepFashion datasets show the efficacy and efficiency of our model. Compared with state-of-the-art methods, our model generates sharper and more realistic images with rich details, while having fewer parameters and faster speed. Furthermore, our generated images can help to alleviate data insufficiency for person re-identification.