Progressive Pose Attention Transfer for Person Image Generation

TL;DR

This paper introduces a progressive pose transfer GAN that generates realistic person images with improved appearance and shape consistency by transferring pose regions sequentially, validated on Market-1501 and DeepFashion datasets.

Contribution

It presents a novel progressive pose transfer network with Pose-Attentional Transfer Blocks that enhances image realism and consistency over previous methods.

Findings

Generated images show better appearance and shape consistency.

The method outperforms previous pose transfer techniques.

It can generate training data for person re-identification.

Abstract

This paper proposes a new generative adversarial network for pose transfer, i.e., transferring the pose of a given person to a target pose. The generator of the network comprises a sequence of Pose-Attentional Transfer Blocks that each transfers certain regions it attends to, generating the person image progressively. Compared with those in previous works, our generated person images possess better appearance consistency and shape consistency with the input images, thus significantly more realistic-looking. The efficacy and efficiency of the proposed network are validated both qualitatively and quantitatively on Market-1501 and DeepFashion. Furthermore, the proposed architecture can generate training images for person re-identification, alleviating data insufficiency. Codes and models are available at: https://github.com/tengteng95/Pose-Transfer.git.