CLASSIFICATION OF FRUITS RIPENESS USING CNN WITH MULTIVARIATE ANALYSIS BY SGD

The classification of fruit ripeness is a critical task in agriculture and food industries, impacting quality control and minimizing waste. This study presents a Convolutional Neural Network (CNN) with VGG16 model for classifying the ripeness of fruits using RGB images and multivariate analysis optimized by Stochastic Gradient Descent (SGD). The process begins with image data preprocessing, where fruits images are cropped and resized to create a standardized dataset. Following this, the dataset is normalized and split into training and testing sets to ensure a balanced evaluation. The CNN model uses the VGG16 architecture for feature extraction, leveraging its deep layers to learn distinguishing features of ripe and rotten fruits. The model is trained using SGD, which optimizes the learning process by updating weights in small, frequent steps, enhancing convergence speed. The classification is performed on three fruit types: bananas, papayas, and oranges, with the final output categorizing the images into six classes: ripe banana, rotten banana, ripe papaya, rotten papaya, ripe orange, and rotten orange. This method achieves effective feature extraction and classification, providing an accurate system for automated fruit ripeness detection. The proposed approach offers potential for real-world applications in agriculture and automated food inspection systems.

Keywords: Fruit Dataset, Deep Learning, Convolution Neural Network, Image Processing Techniques and accuracy.