Real-Time Vegetable Identification and Detection Implementing Yolo

Abstract

Fruits and vegetables are essential sources of nutrition, crucial for preventing chronic diseases like diabetes, cancer, and cardiovascular issues. Maintaining a healthy diet, including the recommended servings of vegetables and fruits, is important for overall well-being. However, many individuals, particularly those unfamiliar with grocery shopping, struggle to distinguish between different types of vegetables due to their similar physical characteristics, leading to confusion and poor dietary choices. Therefore, this project aims to design and develop a real-time vegetable detection and identification system through a mobile application using deep learning and provide healthy recipes based on the identified vegetable. This project used the YOLOv5 object detection algorithm for vegetable detection, utilizing manually captured vegetable images from phones as the dataset. The collected data were organized into a unified folder. The Modified Waterfall model was adopted as the methodology, excluding the maintenance phase, encompassing requirement gathering, design, implementation, and testing. The testing phase demonstrated that the model met all the project’s objectives and successfully identified and detected vegetables with a mAP of 95.7%. All functionality test cases and an accuracy test confirmed that the system effectively resolved the problem. The system was developed as a mobile application to enhance accessibility for target users. Future enhancements could include refining the detection model and incorporating a wider range of vegetables or including other ingredient datasets to broaden its scope.