svip-lab/impersonator
A PyTorch implementation of a GAN framework for synthesizing human images with motion transfer, appearance changes, and novel viewpoints.

This repository provides a PyTorch implementation of the Liquid Warping GAN paper from ICCV 2019. The framework handles three tasks: human motion imitation (reenacting target poses on a source person), appearance transfer (applying clothing/appearance from a reference image), and novel view synthesis (rendering a person from different angles). It uses GANs with body parsing and pose keypoints to achieve these transformations on person images.