Tomato Leaf Disease Classification with Lightweight Federated Learning using Identically Distributed Images
Naresh Kumar Trivedi, Himani Maheshwari, Raj Gaurang Tiwari, Ambuj Kumar Agarwal, Vinay Gautam · 2023
Tomatoes grown and consumed extensively in Asia are an essential source of protein (22% per Tomato) and energy (18% per Tomato). Bacterial, fungal, and other microbial diseases severely impact plant health and productivity for tomato farmers. Manual diagnosis of many diseases is challenging, especially in places that need access to crop protection experts. Automated disease identification is crucial for Tomato leaf protection and reduced crop loss. Unfortunately, no one method of categorizing leaf diseases in tomatoes has proven to be consistently accurate enough for widespread usage. The present study offers a streamlined federated deep learning framework to classify tomato leaf diseases while ensuring the preservation of sensitive data. This framework safeguards client information with a decentralized client-server architecture and verifies federated deep learning models with IID. The researchers piloted the framework with a single and several clients in both conventional and federated learning environments. After extracting features from numerous pre-trained models, AlexNet is selected as baseline because of its superior 99.53% accuracy. Next, IID datasets were put through their paces using federal learning (FL). The extra capabilities of the framework make it the ideal option for early Tomato leaf disease categorization.