Projects
Computer Vision & Machine Learning for Transportation
YOLOv8 Traffic Detection for Dhaka
Deployed a YOLOv8-L object detector on Dhaka traffic imagery using a large, manually labeled dataset. Focused on multimodal traffic (car, bus, CNG, rickshaw, etc.) and built a reusable pipeline for robust detection and analysis.
Technologies: Python, YOLOv8, Computer Vision
Dhaka Corridor Multimodal Traffic Counts from Video
Used a YOLOv8-based detector, line-based tracking, and hand-drawn regions of interest on a corridor video to generate lane-level multimodal traffic counts. Captured flows of pedestrians, rickshaws, bicycles, and vehicles and identified contra-flow movements for corridor-level analysis.
Technologies: Python, YOLOv8, Video Analytics
Road Traffic Accident Severity Prediction
Developed machine learning models to predict accident severity classes (Slight, Serious, Fatal) using an imbalanced accident dataset. Applied techniques such as SMOTE and undersampling, and evaluated models including Random Forest and XGBoost with F1-score and feature importance analysis.
Technologies: Python, Scikit-learn, XGBoost, Data Analytics
Transportation Modelling & Research
Urban Freight Attraction Modelling in Dhaka City
Undergraduate thesis project on establishment-level freight trip attraction in Dhaka. Collected and analysed establishment-based data to model freight attraction in a dense, developing-city context, contributing to better understanding of urban freight patterns.
Focus: Freight generation & attraction, urban logistics, data-driven modelling