Few-Shot Learning for Multiclass Disease Detection in Azadirachta Indica

Plant disease detection is a crucial field where numerous researchers are exploring today. Various deep learning models are employed to detect different diseases in plants. Few-shot learning is one of the emerging techniques for classification tasks from the past few years in agriculture. The study identifies various diseases of Azadirachta indica using deep learning model with a machine learning framework called few-shot learning. The dataset consists of images of leaves and trees categorized into five classes. The model uses a ResNet-50 pre-trained model as the backbone for feature extraction and adds custom classification layers for few-shot learning. This model leverages its efficiency by combining meta-learning and metric-learning approaches. This study focuses on neem leaves as they have more therapeutic value. The work shows promising results in combining meta-learning and metriclearning approaches for few-shot learning.