A Novel Multi-Task Deep Learning for Plant Classification and Disease Diagnosis from Leaf Images

• Published in: 2024 International Conference on Power, Energy and Innovations (ICPEI) pp. 109 - 113
• Main Authors: Sue-Chayapak, Wisuttipong, Suhren, Wiwat, Srikaew, Arthit
• Format: Conference Proceeding
• Language: English
• Published: IEEE 16-10-2024
• Subjects: Accuracy; Agriculture; Convolutional neural networks; Convolutional Neural Networks (CNNs); Deep learning; EfficientNetB0 Architecture; Medical diagnosis; Multi-task deep learning; Multitasking; Plant diseases; Resource management; Technological innovation; Training
• DOI: 10.1109/ICPEI61831.2024.10749297

The analysis of plant diseases using leaf images through deep learning techniques have been attracting great attention. The Convolutional Neural Networks (CNNs) plays an important role in analyzing plant leaf images to accurately identify plant species and diagnose diseases. This paper presents a newly modified of deep learning in plant identification and disease classification. By using the PlantVillage dataset, various plant species leave images and their corresponding disease can be learned simultaneously. The proposed model, called multi-task deep learning for classifying plant species and diagnosing plant diseases concurrently is based on the EfficientNetB0 architecture. This technique can reduce training time while maintain effective classification and diagnosis of plant diseases. The experimental results show that the proposed model can classify plant leaf types and diagnose plant diseases with up to 99.95% of accuracy. This firmly demonstrates the powerfulness of deep learning technology as a significant tool for agriculture in monitoring and managing plant diseases, helping to reduce potential crop losses and improve agricultural productivity.