Built a monocular visual odometry pipeline using FAST corners, KLT tracking, PnP RANSAC, and multi-view triangulation, achieving 100% pose estimation success across 13,000+ frames on four datasets
Fused 6-axis IMU data with visual estimates to recover metric scale (4.2× factor) and validated elevation gain against barometric altitude to within 1.2% error
Collected a custom cablecar dataset in Switzerland and used gyroscope priors to maintain tracking through sustained rotation where pure vision failed
Developed a modular Python application implementing key ML algorithms: SVM, K-Nearest Neighbors (KNN), Linear Regression (Gradient Descent & Normal Equation)
Integrated end-to-end ML pipeline with data loading, preprocessing, and k-fold cross-validation
Implemented GUI (PySimpleGUI) for model parameter selection and training visualisation
Designed reusable modules for cross-validation, ROC curve plotting, and dataset management
Tested and validated all major components (e.g., svm_test.py, roc_curve_test.py)
Organised >2000 LOC across cleanly structured directories: algorithms/, data/, plots/, tests/
Produced comprehensive documentation and reports, including performance analysis and research notes
