Paper Title
Fruit Sense: A System for Automated Fruit Grading and Price Estimation

Abstract
This paper presents FruitSense, an AI-based system for automated multi-fruit quality grading and quality-aware price estimation. The proposed system supports apple, banana, and orange through a modular deep learning pipeline. A MobileNetV2-based fruit classification model first identifies the fruit type from the input image, after which a fruit-specific convolutionalneural network (CNN) performs quality grading based on visual attributes such as freshness and surface defects. The system is evaluated under varying real-world imaging conditions, including changes in lighting, background, and image quality, to assess robustness and generalization. A grade-driven price estimation subsystem employs a supervised machine learning model trained on historical market data and adjusts predicted base prices using quality-dependent scaling factors derived from visual grading. Experimental results demonstrate high classification accuracy across supported fruits, while interpretability analysis using Grad-CAM provides transparency into model decision-making. Comparative evaluation highlights the scalability and practical applicability of the proposed approach, positioning FruitSense as a decision-support tool for AI-assisted agricultural grading and market-oriented quality assessment. Keywords - Deep Learning, Computer Vision, Automated Fruit Grading, Multi-Fruit Classification, Quality-Aware Price Estimation, Agricultural AI