An Electronic Nose Approach in Identifying Ripe Tomato Fruit Using LeNet-5 Structure

Tomatoes, a major Philippine crop, require accurate ripeness assessment for optimal quality. While odor is a valuable indicator, current electronic nose (e-nose) systems lack real-world applications of deep learning. This study investigates the impact of data augmentation and input size on LeNet-5 performance for tomato ripeness classification using an e-nose system. Tomatoes were categorized as ripe, not ripe, or unknown. The LeNet-5 model trained with data augmentation and an input size of 24x8 achieved the highest mean score (0.95) during training. Implemented in the e-nose system, this model classified 43 out of 50 samples correctly (86% accuracy).