This thesis project explores machine learning-based solutions for detecting and classifying seven common strawberry diseases to promote sustainable smart farming in Ecuador. ​ Using advanced computer vision techniques, the study fine-tunes state-of-the-art models like YOLOv10, FR-DETR, and SAM2 on a dataset of 4,900 images collected from Ecuadorian fields and open-source platforms. ​ YOLOv10 achieved a mAP@0.5 of 0.95 and an F1-score of 0.94, demonstrating strong real-time detection capabilities, while FR-DETR attained a comparable mAP@0.5 of 0.961 but exhibited higher computational demands. ​ SAM2 was fine-tuned for precise segmentation of diseased areas, enhancing detection accuracy. ​ The research highlights YOLO's suitability for practical deployment due to its speed and efficiency, while FR-DETR's transformer-based architecture excels in global context modeling for larger diseased regions. ​ Challenges such as dataset imbalance and class-specific detection issues were identified, with recommendations for future improvements through data augmentation and hyperparameter tuning. ​ The findings underscore the potential of AI-driven tools in transitioning Ecuadorian strawberry farming from traditional pesticide-heavy methods to more sustainable practices. ​