Here, you'll find a collection of my projects that demonstrate my skills and passion in artificial
intelligence and machine learning. From data analysis to model development, each project reflects my growth and dedication to the field.
AI-Powered 3D Football Match Reconstruction from Camera Footage
This project focuses on transforming standard football match footage into a detailed 3D tactical representation. Using camera diffusion video as input, we apply advanced AI techniques including Open Pose for accurate multi-player pose estimation and deep learning models to detect key field landmarks. By estimating the homography between the video and a reference layout, we align player positions to the field in real-time. The result is a dynamic 3D replication of the match, offering insights into player movement, positioning, and overall strategy.
Real-Time Player
Trajectory Reconstruction
This project creates a spatio-temporal tactical map to reconstruct player trajectories in real-time using video streams and geometric models, without GPS sensors. The video demonstrates the tracking of Player ID 2 and their trajectory on the field layout.
Tactical Map Generation
from Football Match Footage
This project focuses on creating a tactical map of a football match by tracking players through diffusion camera footage. Using deep learning, I developed a model that identifies key points on the field, matches them with a reference layout, and estimates homography to accurately represent player movement and strategies.
Joint Detection and 3D Mapping Using Stereo Cameras and YOLO
This project uses a stereo camera setup with YOLO to detect joints across synchronized image pairs. By applying triangulation, it reconstructs the 3D positions of joints, then uses homography to align them to a football field layout. The result is a real-time 3D mapping of player posture and movement.
T2 Mapping for Disc Hydration Estimation
This project aims to create a T2 map from multi-echo MRI signals to estimate the relaxation times of intervertebral discs, providing an approximate indication of the hydration levels of the discs.
Knee MRI Segmentation Across Sagittal, Axial, and Coronal Planes
This project involves applying deep learning techniques to segment knee MRIs across all sagittal, axial, and coronal slices, enabling more accurate and comprehensive analysis.
