A Comprehensive Review of Machine Learning and Deep LearningTechniques for Banana Leaf Disease Detection

Abstract

Banana leaf diseases like Panama wilt, Banana Bunchy Top Disease (BBTD), and various fungal infections can seriously hurt crop yields and the farmers’ bottom line. Lately, machine learning and deep learning have stepped up, giving us automated image-based tools that spot these diseases early and help farmers make smarter decisions. In this review, I dig into 29 research studies on banana leaf disease detection. I look at widely used datasets like BananaLSD [1], old-school image processing methods [17], convolutional neural networks (CNNs) [2], [4], hybrid deep learning models [7], [22], segmentation-based approaches [8], [23], and even YOLO for object detection [14].

 

When you stack them up, CNN-based and hybrid models leave traditional machine learning approaches in the dust. Still, there’s a lot to tackle—things like limited datasets, changing environments, and the headaches that come with heavy computation and getting these systems up and running in the real world. Lightweight CNNs get a special spotlight here because they try to strike that tricky balance between accuracy and speed. This review points out where the research still falls short and sketches out where we need to head next if we want banana leaf disease detection to be scalable, efficient, and transparent.