A Comparative Analysis of Few-Shot and Zero-Shot Learning Models with Explainable AI for Crop Disease Recognition Emerging Smart Agriculture

The Rapid incorporation of Artificial Intelligence into the agriculture sector transforms the shape of traditional agriculture. Among various problems in the agriculture sector, plant disease is one of the major factors that can reduce crop production and quality. The main challenge to overcome this situation is to find the disease accurately, as we have a scarcity of data in agricultural research. This paper provides a comparative analysis of different Convolutional Neural Networks (ResNet- 50, EfficientNet, DenseNet121) Models to detect Potato Leaf diseases. The main focus of this research is to implement the Few Shot Learning (Prototypical Network) and Zero Shot Learning (GAN and CLIP) models to get better performance rather than traditional CNN models. In addition to these, Explainable AI (LIME) has been deployed to increase the interpretability of the models. Different performance matrices such as Accuracy, Precision, Recall and F1 Score are applied to evaluate the efficiency of models.