Automated Coffee Leaf Disease Detection Using Convolutional Neural Networks: An EfficientNet-Based Approach
Pratham Kaushik, Pooja Sharma
Abstract
The deep learning system successfully spots and labels coffee leaf diseases at 81% accuracy. The model tested well both during training and validation sessions with minimal risk of wrong learning patterns. The model achieved strong results for detecting “No Disease” and “Miner” cases with high precision and recall but struggled to separate “Phoma” and “Rust” samples because the leaf diseases look similar. The model achieves dependable disease detection in each category based on weighted and non-weighted F1-scale measurements. The model shows promise but needs better ways to detect poorly performing classes. The next steps will focus on building this model through expanded datasets plus better extraction methods and advanced learning methods. This system functions as a dependable coffee leaf disease detector that offers valuable solutions for real-world agricultural management of coffee fields.