Tam kavramsal sinir ağları kullanılarak tarımsal zarar ve hareketlerin tespiti ve sınıflandırılması

This thesis undertakes a comprehensive exploration of the development and implementation of Full-Convolutional Neural Networks (FCNs) for precise agricultural pest and vermin detection and classification. Precision farming, a cornerstone of modern agriculture, relies heavily on effective crop health management, with pest and rodent identification being a critical aspect. FCN, a specialized deep learning technique, exhibits unique capabilities by facilitating precise pest localization within crop images, regardless of input size. To ensure the research's credibility and efficacy, a diverse array of data sources is tapped, encompassing drones, stationary cameras, and handheld devices. High-quality images characterized by superior resolution and ideal lighting conditions are essential for maximizing model performance. Additionally, dataset diversity and rigorous professional-grade annotation processes significantly bolster model robustness. Annotation tools such as Labelbox are harnessed for accurate pest and vermin delineation, employing advanced techniques like bounding boxes and segmentation masks, with multiple reviewers engaged to minimize errors. This thesis not only presents a comprehensive roadmap for developing and deploying FCNs tailored to agricultural pest and vermin detection and classification but also offers invaluable insights into the seamless integration of cutting-edge technology into precision farming practices. The research thereby contributes to the enhancement of crop health management and the promotion of sustainable agricultural approaches, underlining its significance in the contemporary agricultural landscape.