Bridging CNNs and Transformers: A Comparative Study of EfficientNet, ResNet, and ViT for Plant Disease Detection

The research is to study and compare Vision Transformers (ViT), EfficientNet, ResNet Deep learning algorithms for identifying plant diseases. Advanced Neural Networks based model for plant disease diagnosis. In order to facilitate automated illness identification, the work uses the EfficientNet framework for classify photos of Crop leaves into several disease categories. We use a large dataset of augmented photos of several plant diseases that we got from Kaggle. To improve model generalization, the dataset is subjected to preprocessing procedures such downsizing, normalization, and data augmentation. The implementation makes use of the PyTorch and TensorFlow frameworks to guarantee reliable training and assessment. To increase accuracy and decrease overfitting, the model is trained batch-wise utilizing optimization strategies including regularization and adaptive learning rate modifications. The classification efficiency is Examined through a range of metrics and indicators, and the conclusions show that the model can correctly identify plant illnesses. The suggested system is a useful tool for initial illness intervention through delivering an automated and scalable technique for agricultural disease surveillance. The system's use of deep learning reduces the need for human interaction while guaranteeing quick and accurate disease diagnosis, which helps farmers manage their crops more successfully. For wider accessibility and useful application in smart agriculture, future developments might incorporate real-time image capturing and cloud-based deployment.