CV

Name: Lutong Qin
Email: 211843003@njnu.edu.cn
Location: Nanjing, Jiangsu
GitHub Page

Research Interests: Deep Learning, Computer Vision, Sensor Data Recognition

Education

• Master’s in Electronic Information (Control Engineering)
  • Nanjing Normal University (211 Project University)
  • School of Electrical and Automation Engineering
  • Specialization: Machine Learning, Deep Learning
• Bachelor’s in Automation
  • North University of China
  • College of Electrical and Control Engineering
  • Specialization: Automatic Control Theory, Microcontroller Principles, and Applications

Research Achievements

Papers:

• “Towards Better Accuracy-Efficiency Trade-Offs: Dynamic Activity Inference via Mutual Learning from Various Width-Resolution Configurations” (Status: Major Review.)
  Authors: Lutong Qin, Lei Zhang✉, Chaoda Song, Dongzhou Cheng, Shuoyuan Wang, Hao Wu, Aiguo Song
  • Description: This paper proposes an efficient activity recognition framework to address the challenges of deep neural networks under different computational resource constraints. Traditional static neural networks struggle to adapt to real-time changes in computing resources, especially on mobile devices. Our approach emphasizes jointly considering network width and input resolution to capture multi-scale feature representations. In contrast, previous methods either ignore these factors or adjust them separately. Our mutual learning framework balances accuracy and speed at runtime, outperforming individual adjustments to width or resolution. By mutual learning on different tasks, our framework can adapt to different mobile devices without retraining or loading models. Furthermore, this framework is not dependent on specific network designs and can be applied in various scenarios. Research results demonstrate that our approach not only provides neural networks that adapt to changes in computational resources but also improves the performance of a single model, making it a practical solution.

• “MaskCAE: Masked Convolutional AutoEncoder for Self-Supervised Human Activity Recognition” (Status: Published)
  Authors: Dongzhou Cheng, Lei Zhang, Lutong Qin, et al.
  IEEE Journal of Biomedical and Health Informatics. (CCF-C class journal, TOP journal in Zone 1 of Chinese Academy of Sciences, IF: 7.7)

• “Revisiting Large-Kernel CNN Design via Structural Re-Parameterization for Sensor-Based Human Activity Recognition” (Status: Published)
  Authors: Minghui Yao, Lei Zhang, Dongzhou Cheng, Lutong Qin, et al.
  IEEE Sensors Journal. (TOP journal in Zone 2 of Chinese Academy of Sciences, IF: 4.3)

Computer Software Copyrights:

• Lutong Qin, Lei Zhang, Wang Shuaishuai, Cheng Dongzhou, Xiong Ting. Adaptive Activity Recognition System Based on Mutual Learning Framework.
• Lutong Qin, Lei Zhang, Wang Shuaishuai, Cheng Dongzhou, Yang Guangyu, Xiong Ting. Human Pose Recognition System Based on Dynamic Gaussian Kernel.
• Lutong Qin, Lei Zhang, Xiong Ting, Wang Shuaishuai, Cheng Dongzhou, Yang Guangyu. Human Pose Recognition System Based on Dynamic Path Calculation.

Internship Experience

• Jiangsu Fute Information Technology Co., Ltd.
  • October 2022 - January 2023
  • Cross-project: Intelligent Face ID and Smart Garbage Classification Assistance Recognition System based on Neural Networks

Skills

• Proficient in automation and computer-related knowledge.
• Skilled in Python programming language, PyTorch deep learning framework, and tools like NumPy and Pandas.
• Ongoing improvement of mathematical and algorithmic foundations, including data structures, calculus, probability theory and mathematical statistics, matrix theory, machine learning (by Christopher Bishop), and deep learning (cs231n).
• Proficient in using ChatGPT and Claude.
• Proficient in basic Linux operating system commands.

Honors and Awards

• Graduate Second-Class Scholarship, 2021
• Graduate Third-Class Scholarship, 2022
• Silver Prize in the “Shanxi Province Internet + College Student Innovation and Entrepreneurship Competition,” 2019

Certifications

• GET4
• National Computer Level 2
• Mandarin Level 2 (A)

Summary

• During my master’s program, I conducted research on wearable device-based human activity recognition algorithms under the guidance of my advisor. I have the ability to read and write English papers and actively keep up with top conferences such as CVPR, ICCV, NIPS, and carefully analyze paper code. I am skilled in paper drawing techniques and LaTeX paper formatting skills.
• Currently, I am researching dynamic adaptive convolutional neural networks to enable more efficient deployment on mobile devices to cope with real-time changes in computing budgets. I am familiar with models such as SVM, DNN, CNN, Transformer, RNN, LSTM, and more.
• I have strong communication skills and excellent written and verbal expression abilities.
• I have a collaborative spirit, can withstand work pressure, and have a high level of execution.