Adaptive optimisation-based deep learning for detection of cotton leaf disease

Leaf diseases are highly prevalent in cotton and reduce the yield and affect the quality of the fibre. These diseases can only be detected by hand, are time-consuming and unreliable, and most existing machine learning models suffer from poor generalisation and require significant computational resources. To overcome the above limitations, this paper introduces a hybrid DL model, CNNbEP-SBRNN. It uses a CNN to learn the disease features, the Emperor Penguin Optimizer (EPO) to determine the most relevant features, and a Secretary Bird-Optimized RNN to predict diseases in cotton leaves. The Cotton Plant Disease dataset is enhanced and prepared with the utmost attention to promote data consistency. Experimental results show that the proposed method has high accuracy, sensitivity, precision, F1-score, and MCC, and performs better than several existing models. The findings demonstrate that the CNNbEP-SBRNN model is credible, efficient and can be used to automate the detection of cotton leaf diseases in real agricultural settings.