Research

Generative Models & Diffusion

A core strength of the lab: controllable image and video generation, editing, and translation. From StarGAN's multi-domain image translation and high-resolution virtual try-on (VITON-HD) to modern diffusion methods for 360° generation, video control, and structure-aware synthesis, we push both the fidelity and the controllability of generative models.

46 publications 5,756 citations

Selected Publications — 5 most representative of 46

StarGAN: Unified Generative Adversarial Networks for Multi-Domain Image-to-Image Translation

CVPR 2018 3,898 cites Oral

StarGAN: Unified Generative Adversarial Networks for Multi-Domain Image-to-Image Translation

StableVITON: Learning Semantic Correspondence with Latent Diffusion Model for Virtual Try-On

CVPR 2024 235 cites

StableVITON: Learning Semantic Correspondence with Latent Diffusion Model for Virtual Try-On

Paper

VITON-HD: High-Resolution Virtual Try-On via Misalignment-Aware Normalization

CVPR 2021 444 cites

VITON-HD: High-Resolution Virtual Try-On via Misalignment-Aware Normalization

Paper

High-Resolution Virtual Try-On with Misalignment and Occlusion-Handled Conditions

ECCV 2022 212 cites

High-Resolution Virtual Try-On with Misalignment and Occlusion-Handled Conditions

Paper

Reference-based Sketch Image Colorization using Augmented-Self Reference and Dense Semantic Correspondence

CVPR 2020 179 cites

Reference-based Sketch Image Colorization using Augmented-Self Reference and Dense Semantic Correspondence

Paper