Feature Extraction and Classification of Smut Disease in Sugarcane Leaves Using CNN and KNN

This research study presents the evaluation results of a classification model for determining the Damage degree of sugarcane leaf diseases. The performance of the model has been evaluated through Precision, Recall, and F1-Score metrics resulting in significant output overall damage levels More precisely, Damage Degree 1 (which means disease severity of type between 1 and 20<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>) is detected with precision/recall/Fi-Score=93.75% On another hand for Damage Degree 2 (21-40%) it has had a slight down with the metrics all this is at approximately 93.72 and ended to be precision, recall, and F1-Score of 94.06 which got one highest number among others. With 10,030 photos used to train it, the model's total accuracy was 94.01%. The highest values are for the most severe conditions, Damage Class 4 (61-80%) with a predictive accuracy of 94.20%, and those belonging to classed Damage Degree5(81-100%), which have an accuracy over classic Ink-based PTA test at an almost same higher point equal form OCR system around about 94.31%. With 10,030 photos in the dataset, the model's overall accuracy was 94.0179%. These are the support values that give the total sample counts for each damage class from Damage Degree 1 (1920 number of samples) to result in a number-of-samples-from-DD2,110array. They do this by making each class around 19%-21% of the dataset. It is very stable with 98% accuracy across all damage degrees, illustrating that the system can predict different stages of a disease degree. The results of this study demonstrate that the model demonstrates a great capability and consistency in diagnosing sugarcane leaf diseases precisely with a high accuracy rate.