Yuan Li
Researcher at 北京大学
Beijing, China
Summary
🤩
Rockstar🎓
Top SchoolYuan Li is a researcher based in Beijing with eight years of experience at the intersection of video codec hardware design and machine learning-driven vision systems. With a PhD in Electrical and Electronics Engineering from Peking University and a deep background in SoC and encoder architecture, he has transitioned from designing fractional motion estimation and CABAC blocks for video encoder chips to contributing ML engineering work on high-profile vision transformer research. His open-source contributions to the ICCV2021 T2T-ViT repository focus on model visualization and feature-map tooling, highlighting an emphasis on interpretability and adapting visualization across ViT and ResNet families. At Peking University he has combined academic rigor with hands-on system design, moving between postdoctoral research and current research roles that blend theory and application. Colleagues would describe him as a technically versatile engineer who bridges low-level hardware-aware signal processing and modern deep-vision model introspection.
8 years of coding experience
Doctor of Philosophy (Ph.D.), Electrical and Electronics Engineering, Doctor of Philosophy (Ph.D.), Electrical and Electronics Engineering at Peking UniversityBachelor's degree, Electrical and Electronics Engineering, Bachelor's degree, Electrical and Electronics Engineering at South China University of Technology
Chinese
Github Skills (8)
computer-vision10
pytorch10
visualization10
visualizations10
vision-transformer10
ml9
machine-learning9
mle9
Programming languages (2)
Github contributions (5)
yitu-opensource/T2T-ViT
Jan 2021 - Jun 2022
ICCV2021, Tokens-to-Token ViT: Training Vision Transformers from Scratch on ImageNet
Role in this project:
ML Engineer
ML EngineerContributions:1 release, 61 commits, 1 PR in 1 year 4 months
Contributions summary:Yuan contributed significantly to the project by modifying and adding files related to model visualization. They updated visualization code, implemented feature map saving and display, and adapted the code for both Vision Transformer and ResNet models. These changes indicate a focus on understanding and interpreting the model's internal workings, aligning with the project's goal of training and visualizing vision transformers. The user also modified core model files, specifically updating hyperparameters and incorporating various T2T-ViT model configurations.
pytorchimagenetvision-transformertransformerstokens
yuanli2333/yuanli2333.github.io
Jan 2022 - Dec 2024
Contributions:62 pushes, 2 branches in 2 years 11 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.