Sugarcane pest and disease detection consider techniques in precision agriculture that would not only be expected to maintain the crops' health but even improve their yields. This research proposes assessing the various detection algorithms for their efficiency in detecting and classifying sugarcane diseases, including Deep Learning (CNN, ResNet, YOLO), Machine Learning (SVM, Random Forest, K-NN), Image Processing, and Hybrid Approaches. A trend study examining 2020 to 2024 shows that the response to research between these years will sharply increase; deep learning is the methodology because of its high accuracy and real-time detection measures. However, machine learning and hybrid models contribute to good classification and predictive modeling levels. Performance evaluation is based on accuracy, precision, Recall, F1-score, computation time, and Robustness, showing a trade-off in model accuracy and computational efficiency. Dataset size vs. accuracy shows that while increasing data size improves performance in early measures, the returns will diminish beyond a threshold effect and further emphasize the quality and diversity of the dataset. In compliance with the above, such systems ought to be developed in consideration of better model architecture for improved database standardization and optimizations in computation, paving the way for much-enhanced precision farming.
