Real-Time Vegetable Classification for Smart Retail and Agriculture using Edge Impulse

Vegetable categorization is one of the key factors to diminish food loss and optimize agricultural efficiency. The research here emphasizes the creation of an intelligent system based on deep learning methodologies to categorize vegetables. The model utilized here is MobileNetV2 combined with Faster Objects, More Objects (FOMO) for effective object detection. It is trained on five vegetable categories: coconut, green chili, onion, potato, and tomato. The model had an F1-score of 93% on the validation set, precision of 96%, and a recall of 90% for non-background classes. These show the system’s accuracy in correct vegetable classification. Dataset balancing and hyperparameter fine-tuning are some ways in which the performance of the model can be improved further. The scalability of the system and its potential to expand in the future to cover more food varieties offer possibilities for use, in agriculture businesses, retail industry and automated cooking sectors. This document explains how this creative solution was developed and assessed.