Enhanced Multi-Scale Cross-Attention for Person Image Generation

TL;DR

This paper introduces XingGAN, a novel multi-scale cross-attention GAN for person image generation that effectively captures appearance and shape, outperforming existing methods in quality and speed.

Contribution

The paper presents a new cross-attention mechanism, multi-scale blocks, and a densely connected co-attention module, advancing person image generation with improved accuracy and efficiency.

Findings

Outperforms current GAN-based methods in quality.

Achieves comparable results to diffusion-based methods.

Significantly faster training and inference than diffusion models.

Abstract

In this paper, we propose a novel cross-attention-based generative adversarial network (GAN) for the challenging person image generation task. Cross-attention is a novel and intuitive multi-modal fusion method in which an attention/correlation matrix is calculated between two feature maps of different modalities. Specifically, we propose the novel XingGAN (or CrossingGAN), which consists of two generation branches that capture the person's appearance and shape, respectively. Moreover, we propose two novel cross-attention blocks to effectively transfer and update the person's shape and appearance embeddings for mutual improvement. This has not been considered by any other existing GAN-based image generation work. To further learn the long-range correlations between different person poses at different scales and sub-regions, we propose two novel multi-scale cross-attention blocks. To tackle the issue of independent correlation computations within the cross-attention mechanism leading to noisy and ambiguous attention weights, which hinder performance improvements, we propose a module called enhanced attention (EA). Lastly, we introduce a novel densely connected co-attention module to fuse appearance and shape features at different stages effectively. Extensive experiments on two public datasets demonstrate that the proposed method outperforms current GAN-based methods and performs on par with diffusion-based methods. However, our method is significantly faster than diffusion-based methods in both training and inference.