Disease Detection in Plum Using Convolutional Neural Network under True Field Conditions

Area of Science:

• Agricultural Science
• Computer Science
• Artificial Intelligence

Background:

• Crop diseases cause significant global losses in agriculture, impacting food productivity and quality.
• Early detection of plant diseases remains a challenge for farmers, especially in developing nations.
• Mobile computing and AI offer new avenues for developing practical plant disease detection tools.

Purpose of the Study:

• To propose an efficient deep learning framework for detecting plum diseases under real-world field conditions.
• To optimize the framework for deployment on resource-constrained mobile devices.
• To enhance disease detection accuracy and robustness through advanced data handling techniques.

Main Methods:

• Development of a convolutional neural network (CNN)-based disease detection framework.
• Collection of a unique dataset of plum images captured under diverse field conditions.
• Extensive data augmentation and transfer learning using the Inception-v3 model.
• Parameter quantization for optimizing the model for resource-constrained devices.

Main Results:

• The optimized Inception-v3 model achieved over 92% accuracy in classifying healthy and diseased plum fruits and leaves on mobile devices.
• The framework demonstrated robustness in detecting diseases under challenging, real-world field conditions.
• Transfer learning with scale-sensitive models proved effective with augmented datasets.

Conclusions:

• The proposed AI framework offers an efficient and accurate solution for early plum disease detection on mobile devices.
• The study highlights the potential of deep learning and mobile technology to support farmers in disease management.
• The optimized model is suitable for deployment in resource-constrained environments, addressing practical agricultural challenges.