In this thesis, we focused on the recognition of wheat heads using deep learning methods. The primary objective was to develop a model that accurately detects wheat heads, leveraging YOLOv8, an advanced object detection model. This model utilizes modern hardware capabilities to achieve efficient training and prediction. We obtained a comprehensive set of photographs from the Global Wheat Head Dataset, which includes pre-annotated data. Through multiple phases of training and validation, we have developed a well-trained and refined model with optimized weights, ensuring high accuracy and reliability in detecting wheat heads under diverse environmental conditions. We also tested it on images of wheat fields, obtained with a drone.
|
