MD AYNUL ISLAM (叶子)

Summary

A passionate learner, determined to keep growing through every opportunity and challenge. As a computer science researcher, committed to advancing artificial intelligence through computer vision, medical image analysis, deep learning, and vision-language models. Experienced in developing and evaluating deep learning architectures for object detection, image segmentation, anomaly detection, and real-time intelligent systems. Research experience spans the University of Science and Technology of China (USTC) and Southwest University of Science and Technology, with multiple papers published at peer-reviewed conferences. Outside research: volunteering, cricket and football, reading articles and case studies, brainstorming new AI ideas with peers, and talking with people who are passionate about their own fields.

Research Interests

Computer vision; medical image analysis; deep learning; vision-language models; medical image segmentation; electronic health records (EHR); reinforcement learning.

Research Experience

Education

Selected Papers

1. An Edge-Cloud Collaborative Autonomous Driving System Based on Swin-YOLOv11 (IEEE International Conference on Predictive Control of Electrical Drives and Power Electronics)

   Developed a Swin-YOLOv11 edge-cloud collaborative object detection framework to improve detection accuracy and computational efficiency for autonomous driving.

2. HyperSeg-DG: Multi-Scale Hyper-Feature Context for Domain-Generalized Medical Image Segmentation (Bioinformatics, CCF-A; supported by NSFC 62502491 and the USTC Software Youth Fund YN2260080011)

   Developed a hypergraph-enhanced adaptive visual perception framework for medical image segmentation under domain shift, low light, low contrast, and blurred anatomical boundaries.

3. Mamba Nested U-KAN: Region Attention and KAN-Enhanced U-Net++ for Medical Image Segmentation (Under review — ACM Transactions on Computing for Healthcare)

   Proposed a segmentation architecture combining Mamba, Kolmogorov–Arnold Networks (KAN), region attention, and U-Net++ for blurred foreground–background boundaries and difficult medical imaging conditions.

4. FAMA: Frequency-Aware Multi-Level Vision-Language Model Adaptation for Generalized Zero-Shot and Few-Shot Medical Anomaly Detection (Under review — Bioinformatics)

   Proposed frequency-aware multi-level vision-language model adaptation through pixel-level alignment, frequency decomposition, and dual-branch cross-attention, improving AUC by 5.18% on anomaly classification and 3.21% on anomaly segmentation.

5. Lung Nodule Classification with Quantum Neural Networks and Kolmogorov–Arnold Networks (IEEE RAAICON 2025)

   Studied the combined use of quantum neural networks and Kolmogorov–Arnold Networks for lung nodule classification.

6. Integrating YOLOv8 and Vision Transformers: An Enhanced Security Framework for Digital Twins (IEEE RAAICON 2025)

   Developed a real-time security framework combining YOLOv8, vision transformers, DeepSORT, IoT data, and a digital twin for human activity and threat detection, reaching 93.77% accuracy and 98.14% precision on the Real-Life Violence dataset.

Selected Projects

Technical Skills

Deep learning and AI frameworks: PyTorch, TensorFlow, Keras, Transformers, BERT

AI methods: RAG, YOLO, Mamba, Kolmogorov–Arnold Networks, deep reinforcement learning

Tools: Git, GitHub, Anaconda, LaTeX, Microsoft Office

Honors and Awards

• Outstanding Graduate, Class of 2026 — University of Science and Technology of China, for academic and research performance during graduate study.
• 2026 USTC Anniversary Grand Prize — School of Innovation and Entrepreneurship, USTC, for developing USTC.AI.
• Chinese Government Scholarship, 2023 — for graduate study in China.
• Special Scholarship — Southwest University of Science and Technology.
• Outstanding Student Award — Sichuan University of Culture and Arts, for Chinese language study.

Languages