Advanced Deep Learning Technique for Accurate Pest Detection in Medicinal Plants

Medicinal plants are widely used in traditional and modern Ayurveda for treating various diseases. However, pest infestation poses a significant challenge to maintaining their health and productivity. This paper proposes a method for detecting pest-infected medicinal plants, which may assist farmers, researchers, and agricultural professionals in identifying infestations early. The study utilizes three datasets of medicinal plants which are Neem, Aloe Vera, and Ashwagandha, and trains five deep learning models such as Basic CNN, VGG16, ResNet50, VGG32, and a newly proposed CFNET. These models process plant image data and predict pest presence using metrics such as accuracy, precision, recall, F1-score, and training time. The CFNET model, built on the EfficientNetB3 architecture, achieved the highest accuracy of 0.98 on the Neem dataset, 0.94 on Aloe Vera, and 0.96 on Ashwagandha, with a focus on reducing false negatives. The results highlight CFNET's superior performance and efficiency compared to other models. This system demonstrates the potential of deep learning for sustainable pest management in medicinal plant cultivation.