OccLinker: Deflickering Occupancy Networks through Lightweight Spatio-Temporal Correlation

Fengcheng Yu, Haoran Xu, Canming Xia, Ziyang Zong, Guang Tan

• Keywords: vision-based occupancy network, deflickering, spatio-temporal correlation, plug-and-play

**

Ziyang Zong, Guanying Chen, Zhaohuan Zhan, Fengcheng Yu, Guang Tan

• Keywords: floorplan generation, large language model, diffusion model

Parkinson’s severity diagnosis explainable model based on 3D multi-head attention residual network

Jiehui Huang, Lishan Lin, Fengcheng Yu, Xuedong He, Wenhui Song, Jiaying Lin, Zhenchao Tang, Kang Yuan, Yucheng Li, Haofan Huang, Zhong Pei, Wenbiao Xian, Calvin Yu-Chian Chen

• Proposed an effective end-to-end deep learning model for Parkinson’s disease (PD) severity diagnosis based on facial video is proposed. The model can extract features from PD patient videos and give robust and interpretable diagnostic results, effectively judging the severity of PD disease.

HareMQ Simulation implementation of RabbitMQ core functions

• Developed a C++-based message queue system inspired by RabbitMQ, focusing on implementing various message routing methods such as direct, topic, header, and fan-out exchanges to enhance messaging flexibility.
• Engineered the system to support message persistence and high availability, ensuring reliability and service continuity akin to RabbitMQ’s operational standards.
• Integrated multiple client support and was designed to handle scenarios in big data processing, microservice architectures, and real-time data applications, reflecting a comprehensive understanding of distributed systems.

Related technologies: C/C++, C++11, Boost, Protobuf, Muduo, SQLite3, Gtest

Project

boost-search-engine A search engine implementation based on the Boost library

• Developed a specialized search engine for Boost library documentation, enhancing search accuracy and efficiency using C++, Boost libraries, HTML, CSS, and JavaScript.
• Managed the entire search engine construction process from data preprocessing and index creation to query processing and result presentation, ensuring a comprehensive search experience.
• Implemented advanced features such as word frequency analysis, dynamically updated indexes, and prioritized search result sorting, making the system highly effective for developers seeking quick access to technical documents.

Related technologies: C/C+, C++11, STL, Boost, jsoncpp, cppjieba, cpp-httplib, jQuery, ajax

Project

Load-balanced-online-OJ-system An online algorithm question-solving service, similar to LeetCode

• Developed an online coding and evaluation platform similar to LeetCode, enabling users to write, submit, and evaluate code directly in the browser. Supported and Provided instant feedback and detailed error analysis.
• Implemented key features including question browsing, real-time code editing with Ace editor, and automated code submission and judgment. Integrated custom load balancing to optimize server utilization and response times.

Related technologies: STL, Boost standard library, cpp-httplib, ctemplate, jsoncpp, load balancing design, distributed systems, multi process/thread control, MySQL C Connect, Ace editor, ajax

Project

Google-tcmalloc-simulation-implementation

The tcmalloc is a memory management project open sourced by Google. I implemented the core functions in tcmalloc through learning and simulation, and implemented a high-concurrency memory pool to replace the native malloc memory development method. Its design purpose is to adapt to the characteristics of a single server with multiple CPU cores, and to ensure that the performance of the memory allocator will not drop sharply due to multi-core competition. In order to reduce memory usage, objects of the same memory size are concentrated in spans composed of several adjacent pages.

Related technologies: STL, multi-threading, memory management algorithms, lock-free programming

Project

Multi-Pattern-Logging-System

• Engineered a comprehensive logging system in C++ that supports multi-level logging, synchronous and asynchronous modes, and reliable output to various destinations like console and files. Utilized advanced C++ features and design patterns to ensure thread-safe operations and expandability across different logging environments.

Related technologies: class hierarchy design (inheritance and polymorphism), producer consumer model, multi-threading, design pattern (singleton, factory proxy, builder, etc.), GTest

Project

Event-Driven-Pipeline-Communication-System-Framework

• Developed an event-driven communication framework using Epoll, optimizing inter-process communication with a focus on efficient data handling. Created a modular system that supports easy integration and expansion, providing a robust platform for developers to implement and adapt in various environments.

Related technologies: STL, system programming, multiplexing technology, pipe communication

Project