Xintao Alpha
Research Scientist at Tencent
Shenzhen, Guangdong Province, China
Summary
🤩
RockstarXintao Alpha is a research scientist with a decade of hands-on experience in image and video restoration, based at Tencent ARC Lab in Shenzhen. He is a prolific open-source contributor and ML engineer behind widely used projects such as ESRGAN, Real-ESRGAN and GFPGAN, focusing on practical inference, model interoperability, and production-friendly tooling. His work spans low-level CUDA optimizations to high-level model training and deployment, including contributions that enabled CPU testing, checkpoint interpolation, and ONNX export for real-world applications. Known for bridging research and engineering, he consistently adds usability features—like model management, scale support, and robust inference pipelines—that make state-of-the-art restoration methods accessible beyond academia.
10 years of coding experience
Github Skills (43)
pytorch10
text-alignment10
convolutional-neural-networks10
c-language10
python10
dcgan10
machine-learning10
cgan10
detectors10
super-resolution10
mask-rcnn10
detector10
ncnn10
deep-learning10
image-restoration10
Programming languages (4)
Github contributions (5)
xinntao/EDVR
Aug 2020 - Nov 2020
Winning Solution in NTIRE19 Challenges on Video Restoration and Enhancement (CVPR19 Workshops) - Video Restoration with Enhanced Deformable Convolutional Networks. EDVR has been merged into BasicSR and this repo is a mirror of BasicSR.
Role in this project:
ML Engineer
ML EngineerContributions:8 commits, 12 PRs, 68 pushes in 3 months
Contributions summary:Xintao's commits primarily involve modifications to the `basicsr/models/ops/upfirdn2d/src/upfirdn2d_kernel.cu` file, indicating involvement in low-level, CUDA-based operations. Additionally, there are commits that include changes in `stylegan2_arch.py`, `test_face_dfdnet.py`, `stylegan2_model.py`, and `sr_model.py` files, suggesting their role in integrating and modifying model architectures. The changes also indicate a focus on improving the performance of image and video restoration models in this repository.
pytorchmergedrestorationenhancementimage-restoration
xinntao/Real-ESRGAN-ncnn-vulkan
Jul 2021 - Apr 2022
NCNN implementation of Real-ESRGAN. Real-ESRGAN aims at developing Practical Algorithms for General Image Restoration.
Role in this project:Software Engineer
Software EngineerContributions:37 commits, 2 PRs, 14 pushes in 8 months
Contributions summary:Xintao primarily focused on enhancing the `real-esrgan-ncnn-vulkan` project, a tool for image restoration. Their contributions included supporting more models, including those for anime-specific content and newer versions. They also added functionality to create output directories and updated the project's documentation. Furthermore, the user refactored the code for model loading and added support for scale factors, contributing to the project's overall versatility and usability.
amdncnndevelopingvulkanpractical
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.