# A Generative Model of People in Clothing

## TL;DR

This paper introduces a novel image-based generative model for full-body people in clothing, overcoming previous challenges by splitting the process into segmentation and image synthesis, enabling realistic, diverse human images conditioned on pose, shape, and color.

## Contribution

It presents the first fully image-based, differentiable generative model for people in clothing that does not rely on complex rendering or 3D scans.

## Key findings

- Generated realistic images of people with diverse clothing styles
- Model can be conditioned on pose, shape, and color
- Encouraging results suggest data-driven human generation is feasible

## Abstract

We present the first image-based generative model of people in clothing for the full body. We sidestep the commonly used complex graphics rendering pipeline and the need for high-quality 3D scans of dressed people. Instead, we learn generative models from a large image database. The main challenge is to cope with the high variance in human pose, shape and appearance. For this reason, pure image-based approaches have not been considered so far. We show that this challenge can be overcome by splitting the generating process in two parts. First, we learn to generate a semantic segmentation of the body and clothing. Second, we learn a conditional model on the resulting segments that creates realistic images. The full model is differentiable and can be conditioned on pose, shape or color. The result are samples of people in different clothing items and styles. The proposed model can generate entirely new people with realistic clothing. In several experiments we present encouraging results that suggest an entirely data-driven approach to people generation is possible.