Jingyun Liang
PhD Candidate
Zurich, Zurich, Switzerland
Summary
🤩
RockstarJingyun Liang is a PhD candidate at ETH Zurich's Computer Vision Lab with 10 years of experience specializing in image and video restoration. He is an active open-source contributor and ML engineer behind official repositories like SwinIR and VRT, improving model robustness, large-image support, and dataset compatibility. His work on the widely used KAIR toolbox added SwinIR testing, Charbonnier loss support, JPEG artifact reduction, and PSNRB evaluation, reflecting a focus on practical evaluation and deployment. Jingyun combines deep research instincts with hands-on engineering—refactoring testing pipelines and automating model downloads to streamline reproducible experiments. Based in Zurich, he bridges cutting-edge transformer-based restoration research with production-ready tooling that eases adoption across datasets and tasks.
10 years of coding experience
Github Skills (11)
swin-transformer10
computer-vision10
pytorch10
transformer10
machine-learning10
image-restoration10
deep-learning10
python10
super-resolution10
test-automation9
denoising8
Programming languages (8)
Github contributions (5)
JingyunLiang/SwinIR
Aug 2021 - Dec 2022
SwinIR: Image Restoration Using Swin Transformer (official repository)
Role in this project:
ML Engineer
ML EngineerContributions:1 review, 60 commits, 5 PRs in 1 year 3 months
Contributions summary:Jingyun primarily contributed to the core functionality of the SwinIR image restoration model. Their commits focused on modifying the `network_swinir.py` file, which included bug fixes and enhancements related to image size handling and padding. They also incorporated features like automatic model downloading and made the model compatible with KAIR-trained models. Further commits show modifications and additions to the testing scripts and support for features like color image artifact reduction and large image testing.
image-srlow-level-visionswinimage-denoisingswin-transformer
JingyunLiang/VRT
Jan 2022 - Jul 2022
VRT: A Video Restoration Transformer (official repository)
Role in this project:ML Engineer
ML EngineerContributions:1 release, 1 review, 13 commits in 6 months
Contributions summary:Jingyun primarily focuses on updating and testing the video restoration model (VRT). Their commits involve modifications to the testing scripts (`main_test_vrt.py`), including dataset loading, model preparation, and evaluation metrics. The user adjusts data loading and file paths for compatibility with different datasets, and refactors the code to enhance compatibility and efficiency. These changes contribute to the model's evaluation and performance analysis within a video restoration context.
vision-transformervrtrestorationdeblurringimage-restoration
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.