Hi there!
I'm Linli Shi (史林立), also known as Larry among my English-speaking friends, and indexss online. I am currently a third-year student at the University of Birmingham, UK, studying at the School of Computer Science, where I also serve as a TA in Team Project Lesson.
🎓 Education
I am currently pursuing a dual Bachelor of Science and Bachelor of Engineering in Computer Science. My academic journey began at Central South University, China in 2021 and will conclude at the University of Birmingham, UK in 2025. Throughout these four years at CSU and UoB, I have consistently received scholarships. Given the unique nature of my dual-degree program, I am preparing for graduate applications for Fall 2026, considering both Master's and PhD programs.
🔬 Research
I specialize in the subfield of Gaze Estimation within Computer Vision. Currently, I work as an RA at the Intelligent Robotics Lab in the School of Computer Science at the University of Birmingham. I have successfully reproduced several SoTA models for video-based gaze estimation and plan to release them on GitHub after the paper is approved. Additionally, I am interested in the applications of multimodal methods and LLMs in computer vision tasks, and I am actively conducting related research.
For student academic activities, I serve as the Technical Consultant for the CSU Apple Club, where I am responsible for organizing peer academic support and leading open-source development workshops.
💼 Experience
Teaching Assistant (TA), University of Birmingham
Team Project Lesson
- Guided groups of students in collaborative software development projects, offering mentorship on project management, coding best practices, and teamwork.
Research Assistant (RA), Intelligent Robotics Lab
University of Birmingham
- Collaborated with faculty and postgraduate researchers on cutting-edge projects in computer vision and robotics.
⭐ Projects
🤖 oh-my-vgaze: Focus on transforming SoTA gaze estimation models from top-tier conferences into video-based versions.
- Adapted and implemented several prominent models, including EFE, FullFace, GazeNet, GazeTR, Gaze360, EyeNet, and FaceNet, for video-based gaze estimation, leveraging temporal information to improve performance on video data.
- Conducted comprehensive benchmark tests on well-known datasets EVE and EyeDiap, analyzing the effectiveness of video-based model transformations compared to the original static versions.
- Assessed model performance improvements by incorporating sequential frame analysis, resulting in more robust gaze tracking under real-world video scenarios, with the results to be published on GitHub after paper acceptance.
🤖 Point Distribution Alignment Loss (PDA Loss): A general video-based gaze estimation sequence optimization loss function.
- By leveraging sequential information, the predicted gaze trajectory is aligned with the actual trajectory, increasing the accuracy of video-based gaze estimation.
- The Procrustes distance of trajectory decreased by 0.02.
☕ Backseat: a novel application designed to facilitate music sharing and discovery between users.
- Mainly responsible for the vertical slicing of the Leaderboard and Album details.
- Innovatively implemented dynamic loading from scratch, combined with Angular's page location identification, step-by-step loading data from the database, reducing server load.
- Deploy Gitlab CI/CD pipeline, team collaboration follows agile development mode.
📟 Vanilla Firewall: a robust server-client system designed to manage and enforce firewall rules effectively.
- Designed and implemented a multi-threaded server using socket programming to efficiently manage firewall rules and respond to client queries, ensuring concurrent request handling. Written in C.
- Created a client application that communicates seamlessly with the server, allowing for operations like adding, deleting, and verifying firewall rules, and effectively listing stored rules.
- Emphasized robust error handling and memory management, ensuring all client-server interactions are accurate, while avoiding memory leaks and improving long-term server reliability.
🎉 Hobbies
- 💀 Arch Linux User 💀 and 🌼 proud owner of the full Apple ecosystem 🌼
- 🎻 Classical music enthusiasts
- ⛰️ Hiking enthusiast
- ⚖️ Keyboard Politician
- 📡 NAS Networking Enthusiast
- 👾 League of Legends of course
贡献者
Linli Shi
jackiexiao