Classification of Strawberry Ripeness Stages Using Deep Learning

Abstract

The issues related with the labor-intensive, inflexible, and error-prone manual classification of strawberries in supply chain are the focus of this project. The objective of the project is to accurately classify different stages of strawberry ripeness based on color by creating an effective deep learning system using MATLAB. The procedure is gathering a sizable dataset of images of strawberries taken at different stages of ripeness, pre-processing the images, and then using the image processing toolbox in MATLAB to extract relevant features. The pre-processed dataset was used to train convolutional neural network (CNN) models, such as CNN in MATLAB and YOLOv5 in Google Colab. The mean average precision (mAP) of 0.922 (50%), recall of 0.939, and precision of 0.887 were all attained by the YOLOv5 mode which is equivalent to 92.2% accuracy. The CNN model trained over several epochs with up to 100% accuracy. In the end, this automated system increases agricultural productivity by minimizing errors and manual labor while guaranteeing product quality and consistency throughout the strawberry supply chain.