Deep Ensemble Learning For Three-Stage Maturity Detection Of Indigenous Fruits

Abstract

Accurate identification of stages of maturity in fruits plays a critical role in agricultural quality control and post harvest management. This study proposes an automated approach based on image data for classifying fruit maturity into three categories—Unripe, Fresh, and Rotten. The system employs Convolutional Neural Networks (CNNs) trained on a selected and elevated dataset comprising multiple fruit types. We have combined predictions from three distinct models that were trained independently using ensemble method that averages their output to increase reliability. Streamlit a web based interface enables users to submit images of fruits and get the maturity classification results. Experimental findings indicate that the ensemble model achieves high accuracy, offering a minimally invasive, efficient, productive and scalable solution for fruit quality assessment in agricultural industries.