CLASSIFICATION OF FRESH AND ROTTEN FRUITS USING DIFFERENT CNN MODELS

Abstract

Fruit freshness automated classification is crucial to the agricultural sector. In the traditional procedure, a human being grades the fruit. Additionally, this process is labor-intensive, time-consuming, and ineffective. Additionally, it raises production costs. Therefore, a quick, precise, and automated system that may lessen human effort, enhance production, and decrease manufacturing time and cost is needed for industrial applications. The deep learning- based model for classifying fruit freshness is used in the current work. Various Convolution Neural Network (CNN) models are proposed, and they are implemented using the publicly available "fruit fresh and rotten for classification" kaggle dataset. Three fresh fruit varieties (Apple, Banana, and Oranges) and their rotting category are employed in an experiment using the dataset. From the given fruit photos, traits or attributes are extracted using a CNN model based on deep learning. The input photos are then divided into fresh and rotting categories by a softmax method. The classification of fresh and rotten fruits uses a variety of CNN models, including Resnet50 (50 Layers), InceptionV3 (48 Layers), and VGG16 (16 Layers). The proposed various CNN models accurately and efficiently evaluate the dataset. Later, the accuracy of the proposed CNN models is compared and the highest accuracy among the three CNN models is identified. In this way, the best accuracy CNN models will be identified for classifying the fresh and rotten fruits. KEYWORDS: Deep learning, CNN model, Inception V3, Resnet50, VGG16.