Controllable Human Image Generation with Personalized Multi-Garments

TL;DR

This paper introduces BootComp, a diffusion-based framework for controllable human image generation with multiple garments, leveraging synthetic data creation and filtering to improve quality and applicability in fashion-related tasks.

Contribution

The paper proposes a novel data generation pipeline and a diffusion model with dual denoising paths for high-quality, controllable human image synthesis using multiple garments.

Findings

Effective synthetic dataset construction for garment images.

High-fidelity human image generation with multiple garments.

Versatile application in virtual try-on and pose control.

Abstract

We present BootComp, a novel framework based on text-to-image diffusion models for controllable human image generation with multiple reference garments. Here, the main bottleneck is data acquisition for training: collecting a large-scale dataset of high-quality reference garment images per human subject is quite challenging, i.e., ideally, one needs to manually gather every single garment photograph worn by each human. To address this, we propose a data generation pipeline to construct a large synthetic dataset, consisting of human and multiple-garment pairs, by introducing a model to extract any reference garment images from each human image. To ensure data quality, we also propose a filtering strategy to remove undesirable generated data based on measuring perceptual similarities between the garment presented in human image and extracted garment. Finally, by utilizing the constructed synthetic dataset, we train a diffusion model having two parallel denoising paths that use multiple garment images as conditions to generate human images while preserving their fine-grained details. We further show the wide-applicability of our framework by adapting it to different types of reference-based generation in the fashion domain, including virtual try-on, and controllable human image generation with other conditions, e.g., pose, face, etc.