An Enhanced Hybrid Machine Learning Model for Plant Disease Detection and Classification

Mara Macelaru, Petrica C. Pop, Rares Chiuzbaian, Norbert Kovacs

Abstract: Timely and precise detection of plant diseases plays a crucial role in ensuring good agricultural productivity and food security. Conventional methods of disease detection frequently depend on manual inspection, which may be time-consuming and susceptible to errors. In our paper, we develop an enhanced hybrid machine learning (ML) based model that combines Bayesian Convolutional Neural Networks (B-CNNs) for feature extraction with Gaussian Naïve Bayes (GNB) classification for final decision-making. In addition, we performed various data augmentation methods to strengthen the diversity of the training data and to improve its generalization. Our proposed hybrid ML-based model was trained and validated on the PlantVillage dataset. The performance metrics obtained were impressive, proving that it is highly competitive against existing state-of-the-art solution approaches and demonstrating the high potential of our hybrid ML-based model for real-world applications in smart agriculture.

External IDs:dblp:conf/hais/MacelaruPCK25