Projects
A showcase of my work in machine learning, deep learning, and computer vision for real-world applications.
Downhill Event Detection
Processed 300k+ sensor samples. Achieved 80.6% classification accuracy with Weighted-KNN. Addressed class imbalance with bootstrapping technique using synthetic oversampling via the covariance matrix.
Automatic Control of Anesthesia: Data Analysis and Optimization with Machine Learning
Developed a machine learning-enhanced system within the ACTIVA Project (activa.unibs.it), to predict BIS values and optimize intravenous anesthesia delivery, overcoming traditional closed-loop limitations and enhancing patient safety.
Action Recognition System
Enhanced a pose estimation pipeline using OpenMMLab, achieving 89% accuracy on a custom dataset. Enabled real-time inference on macOS hardware, overcoming previous compatibility challenges. Designed for human-robot collaboration applications.
Industrial Vision System
Designed a liquid level detection system reducing manual inspections by 70%. Implemented adaptive thresholding for varying lighting conditions.