Li Xuhong
Researcher at 百度
Beijing, France
Summary
👤
Senior🎓
Top SchoolLi Xuhong is a researcher and machine learning engineer with a decade of experience specializing in interpretable deep learning, transfer learning, and large-scale computer vision, currently based in Beijing and affiliated with Baidu. He holds advanced training in computer science and applied mathematics from Beihang University and completed a PhD focused on transfer learning and visual perception for autonomous driving at UTC. His open-source contributions to PaddlePaddle/PaddleX notably improved interpretability tooling—enhancing NormLIME/LIME implementations, visualizations, and result management—bridging research ideas with production-ready code. Earlier work spans stereo vision, lane detection, human pose estimation, and practical image preprocessing for OCR, reflecting a strong applied-research to engineering trajectory. A fan of coding and basketball, he combines rigorous academic depth with pragmatic engineering, often addressing edge cases and usability in model interpretability that many research-focused practitioners overlook.
10 years of coding experience
Master's degree, Computer Science, Master's degree, Computer Science at 北京航空航天大学
Stackoverflow
2,348reputation
259kreached
56answers
11questions
tensorflow
top-5%
Github Skills (13)
paddlepaddle10
computer-vision10
machine-learning10
interpretation10
python10
scikit-learn9
tensorflow9
scikit9
matplotlib8
neural-network6
deep-learning6
queue6
image6
Programming languages (4)
Github contributions (5)
PaddlePaddle/PaddleX
May 2020 - Jun 2020
All-in-One Development Tool based on PaddlePaddle（飞桨低代码开发工具）
Role in this project:
ML Engineer
ML EngineerContributions:6 commits, 3 PRs, 1 comment in 5 days
Contributions summary:Li primarily contributed to the interpretability features within the PaddleX framework, specifically focusing on the NormLIME and LIME algorithms. Their work involved updating and modifying the `interpretation_algorithms.py`, `normlime_base.py`, and `lime_base.py` files, suggesting improvements to the visualization, handling edge cases, and refining the overall functionality of these interpretability methods. Furthermore, they improved the file naming structure for the interpreted results and removed unused arguments.
resnetpaddletensorflowclassificationend-to-end
PaddlePaddle/InterpretDL
Jul 2020 - Jan 2023
InterpretDL: Interpretation of Deep Learning Models，基于『飞桨』的模型可解释性算法库。
Contributions:13 releases, 8 reviews, 285 commits in 2 years 6 months
pytorchmodel-interpretationvision-transformerexplanationspaddlepaddle
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.