Maturity detection for Sevillian-style table olives based on deep learning methods

This thesis explores the application of computer vision and deep learning techniques for olive maturity detection, aiming to optimize harvest timing in olive cultivation. The research focuses on developing a system that can accurately classify olives into five maturity stages using RGB images. Two popular object detection models, YOLO and Faster R-CNN, are employed and trained on a dataset of olive images captured throughout the growing season. The study investigates the models' ability to distinguish between different maturity levels, with particular emphasis on identifying the crucial intermediate stages that indicate optimal harvesting time. The research addresses challenges such as visual similarities between maturity stages and dataset imbalances caused by the rapid maturation process. This work seeks to contribute to the field of precision agriculture by providing olive growers with a tool to enhance crop quality and yield through more informed harvest timing decisions.