Richard Zhang

Richard Zhang is a Principal Scientist at Adobe Research in San Francisco, where his computer vision work on image synthesis builds on the colorization method behind Photoshop's Colorize tool and the LPIPS perceptual metric. He published Colorful Image Colorization in 2016 with Phillip Isola and his UC Berkeley advisor Alexei Efros, coauthored LPIPS in 2018, and joined Adobe Research full-time in 2018. Later work included anti-aliased convolutional networks in 2019, the text-to-image GAN GigaGAN in 2023, and the one-step diffusion method Distribution Matching Distillation in 2024. MIT Technology Review named him one of its 35 Innovators Under 35 in 2023.

Now at: Adobe

Timeline

2016

Colorful Image Colorization

He publishes Colorful Image Colorization in 2016 with Phillip Isola and his UC Berkeley PhD advisor Alexei Efros. The convolutional network turns grayscale photographs into color images and fools human judges in 32% of trials, and the work later becomes the Colorize tool in Adobe Photoshop.
2017

Adobe internship and fellowship

In 2017, he interns at Adobe Research and receives the Adobe Fellowship Award.
2018

LPIPS and Adobe Research

In 2018, he coauthors The Unreasonable Effectiveness of Deep Features as a Perceptual Metric, accepted to CVPR, with Phillip Isola, Alexei Efros, Eli Shechtman and Oliver Wang. The paper introduces a dataset of human perceptual similarity judgments, and the resulting metric is known as LPIPS. After his PhD at UC Berkeley, he joins Adobe Research full-time in 2018.
2019

Anti-aliased convolutional networks

In 2019, he publishes Making Convolutional Networks Shift-Invariant Again, a single-author paper accepted to ICML. It shows that anti-aliasing before downsampling raises ImageNet accuracy for architectures such as ResNet, DenseNet and MobileNet.
2023

GigaGAN for text-to-image

In 2023, he coauthors GigaGAN, a GAN architecture for text-to-image synthesis, with Minguk Kang, Jun-Yan Zhu, Taesung Park and others for CVPR. The model synthesizes a 512px image in 0.13 seconds and a 16-megapixel image in 3.66 seconds.
2023

Named an Innovator Under 35

MIT Technology Review names him one of its 35 Innovators Under 35 in 2023. Adobe credits him with forensic tools that let anyone detect fake images, and with tools that increase transparency about the data behind generated images.
2024

One-step diffusion with DMD

In 2024, he coauthors Distribution Matching Distillation, a CVPR paper with Tianwei Yin, Michaël Gharbi, Taesung Park and colleagues. The method transforms a diffusion model into a one-step image generator with minimal impact on image quality.
As of October 2026

Principal Scientist at Adobe

He is a Principal Scientist at Adobe Research in San Francisco, where his research is in improving generative AI with speed, quality and controllability. In September 2026, three papers he coauthors, TPIPS, UNITE and RAEv2, are accepted to NeurIPS 2026.