Research & Publications
My academic contributions to AI/ML, computer vision, and efficient deep learning
Publications
Research Areas
Presentations
Ongoing Projects
Research Interests
Efficient Deep Learning
Model compression, pruning, knowledge distillation, and quantization for edge deployment
Computer Vision
Object detection, image classification, OCR, and real-time video processing
Natural Language Processing
Bengali NLP, text recognition, and language understanding
MLOps & Deployment
Production ML systems, model serving, monitoring, and CI/CD for ML
Publications
Real-Time Object Detection Optimization for Edge Devices Using TensorRT
This paper presents novel optimization techniques for deploying YOLOv8 models on resource-constrained edge devices like NVIDIA Jetson Nano. We achieved 40% FPS improvement while maintaining 97.5% mAP through strategic model pruning, quantization, and TensorRT optimization.
Ensemble OCR Approach for Bengali Character Recognition in License Plates
A comprehensive study on combining multiple OCR engines (EasyOCR, Tesseract, PaddleOCR) with a voting mechanism for improved Bengali text recognition accuracy. The ensemble approach achieved 15% higher accuracy than individual models on the Bengali license plate dataset.
Efficient Vision Transformers for Low-Resource Deployment
Investigating novel architectures for efficient vision transformers that can achieve state-of-the-art performance on image classification tasks while being deployable on mobile and edge devices with limited computational resources.
Coming Soon
Talks & Workshops
Introduction to Deep Learning with PyTorch
EWU AI/ML Study Group
May 2023
Introductory session on neural network fundamentals and hands-on implementation with PyTorch for beginners.