An Efficient Approach for Potato Leaf Disease Classification Using Cascaded CNN-Transformers

This study introduces an innovative deep-learning method for automatically identifying and categorizing diseases in potato leaves. Early and accurate disease detection is crucial for mitigating crop losses and ensuring potato cultivation sustainability. The proposed method leverages a hybrid model that combines the strengths of ResNet50V2 for robust feature extraction and a pre-trained Enhanced Vision Transformer (ViT) for superior feature representation learning from potato leaf images. This cascaded architecture aims to achieve exceptional performance in disease classification tasks. Extensive experimentation demonstrates the efficacy of the proposed approach in detecting various potato leaf diseases. The successful implementation of this model paves the way for prompt disease detection and efficient potato crop management, ultimately contributing to agricultural sustainability and safeguarding global potato yields.