AI People Wiki

Builders & open source

Jun-Yan Zhu

Michael B. Donohue Associate Professor of Computer Science and Robotics, Carnegie Mellon University

Jun-Yan Zhu is the Michael B. Donohue Associate Professor of Computer Science and Robotics in Carnegie Mellon University's School of Computer Science, where Zhu leads the Generative Intelligence Lab and works on generative models for computer vision, graphics, and computational photography. Zhu is also co-founder and Chief Scientist of Nunchux AI. Before joining CMU, Zhu was a Research Scientist at Adobe Research and a postdoc at MIT CSAIL. Zhu is first author of the CycleGAN paper on unpaired image-to-image translation and co-author of pix2pix, and co-wrote the BSD-licensed PyTorch implementation of both with Taesung Park.123456

At a glance

Current roles
Michael B. Donohue Associate Professor of Computer Science and Robotics +1 more
Sector
University
Education
Carnegie Mellon, MIT, UC Berkeley
Work listed
1
Honors
7
Sources cited
7
Profile reviewed

Career

Zhu completed a PhD in computer science at UC Berkeley from 2013 to 2017, advised by Alexei A. Efros, with a thesis on learning to synthesize and manipulate natural images. Zhu then held a postdoctoral position at MIT CSAIL with William T. Freeman, Joshua Tenenbaum and Antonio Torralba, and worked as a research scientist in Adobe Research's Creative Intelligence Lab.7

Zhu joined Carnegie Mellon University in 2020 as the Michael B. Donohue Assistant Professor of Computer Science and Robotics and became associate professor in 2026, directing the Generative Intelligence Lab. Zhu has taught Learning-based Image Synthesis and Visual Learning and Recognition at CMU, and is co-founder and Chief Scientist of Nunchux AI, which works on multimodal inference for image, video and world models.712

Research

Zhu's research concerns generative models for computer vision, graphics and computational photography. The pix2pix paper (2016) used conditional adversarial networks for paired image-to-image translation, and CycleGAN (2017), with Zhu as first author, learned translation between image collections without paired examples; both are implemented in a BSD-licensed PyTorch repository.4356

The Generative Intelligence Lab studies how creators can control, customize and rewrite generative models, how creators can opt in or out of training data and receive credit, and how synthetic data can help computer vision and robotics. Zhu's homepage notes that work such as iGAN, CycleGAN, Swapping Autoencoder, Custom Diffusion and GigaGAN contributed to features in Adobe Photoshop and Firefly, and that SPADE underlies NVIDIA Canvas.12

Recent work includes SVDQuant for 4-bit diffusion models and a method for generating physically stable brick structures from text, which received the ICCV 2025 Best Paper Award.71

Work and contributions

  • CycleGAN and pix2pix in PyTorch Open-source code

    First author of the 2017 CycleGAN paper (with Taesung Park, Phillip Isola, and Alexei A. Efros) and second author of the 2016 pix2pix paper. The README states that the PyTorch code was written by Zhu and Taesung Park, and the repository's BSD-style license names both as copyright holders.3456

Timeline

  1. Began a PhD at UC Berkeley advised by Alexei A. Efros.7
  2. Co-authored pix2pix, image-to-image translation with conditional adversarial networks.4
  3. Posted the CycleGAN paper on unpaired image-to-image translation as first author.3
  4. Began a postdoc at MIT CSAIL with William T. Freeman, Joshua Tenenbaum and Antonio Torralba.7
  5. Joined Adobe Research as a research scientist in the Creative Intelligence Lab.7
  6. Joined Carnegie Mellon University as Michael B. Donohue Assistant Professor of Computer Science and Robotics.7
  7. Co-authored the brick-structure generation paper that won the ICCV 2025 Best Paper Award (Marr Prize).7
  8. Became Michael B. Donohue Associate Professor and director of the Generative Intelligence Lab at CMU.7

Other documented roles

Roles the sources document that are not confirmed as current.

  • Research Scientist, Adobe Research1
  • Postdoc, working with William T. Freeman, Josh Tenenbaum, and Antonio Torralba, MIT CSAIL1

Honors and recognition

  • ACM SIGGRAPH Outstanding Doctoral Dissertation Award · ACM SIGGRAPH, 20187
  • David J. Sakrison Memorial Prize for Outstanding Doctoral Research · UC Berkeley EECS, 20187
  • Packard Fellowship for Science and Engineering · David and Lucile Packard Foundation, 20237
  • NSF CAREER Award · U.S. National Science Foundation, 20237
  • ICRA Best Paper on Human-Robot Interaction · IEEE International Conference on Robotics and Automation (ICRA), 20247
  • ICCV Best Paper Award (Marr Prize) · International Conference on Computer Vision (ICCV), 20257
  • Sloan Research Fellowship · Alfred P. Sloan Foundation, 20267

Where they work

Suggest a correction or removal

Profiles cover public professional work only, and each statement cites a source. If something is out of date, missing or wrong, tell us and link a source. People can also ask for their profile to be corrected or removed.

References

7 sources · profile reviewed Oct 21, 2026 · data updated Oct 21, 2026

  1. Jun-Yan Zhu — homepage (Carnegie Mellon University)Jun-Yan Zhu (Carnegie Mellon University) · Official page · accessed Oct 21, 2026
  2. Generative Intelligence Lab — Carnegie Mellon UniversityCarnegie Mellon University · Official page · accessed Oct 21, 2026
  3. Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks (arXiv 1703.10593)arXiv · Paper · published Mar 30, 2017 · accessed Oct 21, 2026
  4. Image-to-Image Translation with Conditional Adversarial Networks (arXiv 1611.07004)arXiv · Paper · published Nov 21, 2016 · accessed Oct 21, 2026
  5. junyanz/pytorch-CycleGAN-and-pix2pix READMEjunyanz (GitHub repository) · Repository · accessed Oct 21, 2026
  6. junyanz/pytorch-CycleGAN-and-pix2pix LICENSE (BSD-style)junyanz (GitHub repository) · License · accessed Oct 21, 2026
  7. Jun-Yan Zhu — Curriculum VitaeJun-Yan Zhu (Carnegie Mellon University) · Official page · accessed Oct 21, 2026