Zisheng Cao is an engineer with 11 years of experience specializing in ISP design, camera systems, computational photography and digital filmmaking, currently working at DJI in Hong Kong. He combines deep academic training—a PhD-level focus from the University of Hong Kong and earlier degrees from Tsinghua—with practical R&D experience across Philips Research and ZTE in wireless and imaging domains. At DJI he has applied algorithmic and systems expertise to real-world camera pipelines, and his open-source work includes implementing high-performance CUDA trilinear interpolation for image-adaptive 3D LUTs, showing a focus on optimizing ML-enabled photo enhancement for real-time use. Known for bridging research and product engineering, he brings both low-level performance tuning and end-to-end imaging system design to complex camera problems.
Learning Image-adaptive 3D Lookup Tables for High Performance Photo Enhancement in Real-time
Role in this project:
ML Engineer
Contributions:1 review, 80 commits, 6 PRs in 2 years 5 months
Contributions summary:Zisheng primarily contributed to the project by implementing and refining a trilinear interpolation kernel using CUDA. Their work involved the development of both forward and backward passes for the interpolation, crucial for training and utilizing 3D lookup tables. This suggests a focus on optimizing the performance of image processing tasks within the project, and potentially integrating it within a deep learning pipeline. The user also updated the project's datasets and model parameters, alongside adding a demo program and a visualization tool.
Contributions:79 commits, 77 pushes, 6 comments in 1 year 4 months
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.