SMART GRADING OF WATERMELON SWEETNESS USING IMAGE PROCESSING AND MACHINE LEARNING

The project “Smart Grading of Watermelon Sweetness Using YOLOv8 and YOLOv9 aims to develop an automated system for classifying watermelons as sweet or unsweet using image-based analysis. Manual evaluation of sweetness is often inconsistent and subjective, whereas modern deep learning object detection models can detect subtle visual patterns that correlate with internal fruit quality. This system leverages image preprocessing techniques combined with YOLOv8 and YOLOv9 to detect watermelons in images and classify their sweetness level. The dataset used for this project is collected from Roboflow and contains labeled watermelon images categorized by sweetness. Images are preprocessed to normalize size, reduce noise, and enhance visual features. YOLOv8 and YOLOv9 models are trained to detect individual watermelons and classify them as sweet or unsweet with high accuracy.

A web-based interface developed using HTML, CSS, and JavaScript allows users to upload images, while a Flask backend in Python handles prediction and communicates results to the interface. The system reduces errors associated with manual grading and ensures consistent, objective classification results. Performance evaluation of YOLOv8 and YOLOv9 is conducted to identify the most accurate and efficient model for sweetness detection. This approach demonstrates how state-of-the-art object detection models can be applied to automate fruit quality assessment, and the framework can be extended to classify other fruits or agricultural products based on visual features.

Keywords: Watermelon, YOLOv8, YOLOv9, Image Processing, Sweetness Classification, Deep Learning, Python, Flask, Object Detection