Deep transfer learning for comprehensive diagnosis of cotton leaf pathologies

The cotton sector has recently encountered various obstacles, and traditional methods persist in the identification of cotton leaf diseases. This study has established an automated approach for diagnosing cotton leaf blast disease via deep learning methodologies and image processing. The research included deep learning architectures like Convolutional Neural Network, InceptionV3, ResNet50, VGG16, VGG19, and Xception. The extensive collection consists of over 4200 images, including around 3000 depicting cotton leaf blight and 1200 representing healthy leaves. The results demonstrated that the Convolutional Neural Network models InceptionV3, ResNet50, VGG16, and VGG19 attained final validation accuracies of 92.92 %, 64.1 %, 96.81 %, 95.42 %, and 95.97 %, respectively. The ResNet50 approach has exhibited greater accuracy than previous models, whereas the VGG19 model has achieved the second-highest accuracy. This research enhances precision agriculture by delivering a reliable and precise automated approach for predicting cotton diseases. Subsequent inquiries have been undertaken to enhance the precision and efficacy of deep learning models by the incorporation of cutting-edge technologies, including ResNet50, RegNet, EfficientNetB, and Vision Transformers. This study has resulted in a significant enhancement of cotton leaf diseases through identification with the model of surpassing existing leading methodologies in accuracy, complexity, and inference speed. Thus, the creation of these reliable and precise automated diagnostic tools for cotton leaf diseases markedly enhances precision agriculture. The current investigation could equip farmers with a dependable and effective method to detect and mitigate cotton leaf diseases prior to inflicting significant harm on cotton crops.