COMPARATIVE ANALYSIS OF MACHINE LEARNING METHODS FOR RAPID DETECTION OF VEGETABLE DISEASE USING COMPUTER VISION

Abstract

The timely identification of vegetable diseases is critical for food security across the globe and minimizing agricultural losses. This study compares the performances of four deep learning models – CNN, ResNet50, EfficientNetB0, and MobileNetV2 – by detecting diseases on potatoes, tomatoes, and carrots using computer vision. To determine their effectiveness, these models were trained and tested on a customized dataset and assessed on accuracy and inference time. The maximum accuracy was achieved by the CNN model, according to results. MobileNetV2 is superb; when considering both performance and extreme speed, it stands out for real-time applications. This study exemplifies how lightweight deep learning models can allow for efficient vegetable crop disease detection.