AN AI-DRIVEN APPROACH FOR THE DETECTION AND CLASSIFICATION OF COTTON LEAF DISEASES

Abstract

Cotton, one of the biggest cash crops, is essential to the textile and agricultural sectors. Cotton, on the other hand, is extremely susceptible to several leaf diseases, which impact production and fiber quality. To help in the effective management and timely treatment of the plant diseases, the information on the presence of the diseases in the early stages of occurrence and the correct time of the occurrence is very important. Traditional diagnosis is usually time-consuming, labor intensive, and is less reliable in the field. For that purpose, the current study proposes a deep learning model (ResNet50) architecture for the automatic detection and classification of cotton leaf disease. The SAR-CLD-2024 cotton leaf dataset, which consists of 2,137 original field photos that were enlarged to about 7,000 images through data augmentation and covers seven classes Bacterial Blight, Curl Virus, Healthy Leaf, Herbicide Growth Damage, Leaf Hopper Jassids, Leaf Redding, and Leaf Variegation is used to train and assess the model. The proposed pipeline integrates systematic data cleaning, image preprocessing, augmentation, and transfer learning with fine-tuning of ResNet50's pretrained ImageNet weights to adapt the network to the agricultural. Experimental results show that the proposed model achieves an overall accuracy of 94.60%, with a weighted precision of 93.60%, recall of 92.60%, and F1-score of 93.10% across all seven-disease categories. The results show the classification of cotton leaf disease can be done with reliable and efficient performance by ResNet50, and can be applied to smart agriculture through automatic monitoring of the disease.

Keywords : Cotton leaf disease, Deep Learning, ResNet50, Convolutional Neural Network, Smart Agriculture.