Classification of Capsicum Varieties Using Color Analysis with Convolutional Neural Network

Abstract

Paprika (Capsicum annuum L.) is a high-value horticultural commodity widely consumed for its nutritional content and vibrant color variations. In the agricultural industry, classifying paprika varieties based on color is crucial for ensuring product quality and optimizing sorting processes. This study developed an automated classification system for three main paprika varieties—red, green, and yellow—using the Convolutional Neural Network (CNN) method. The dataset consisted of 1,820 images sourced from Kaggle, with data split into 60% for training and 40% for validation. Preprocessing steps included resizing images, normalizing pixel values to the range [0,1], and data augmentation techniques such as rotation, flipping, and brightness adjustments to enhance dataset diversity and reduce the risk of overfitting. The CNN model was designed with key layers, including convolutional, pooling, and fully connected layers, optimized using the Adam algorithm and categorical cross-entropy loss function. The training results showed an accuracy of 99.9% on the training data and 92% on the testing data, with an average processing time of 64 seconds per image and a maximum of 78 seconds, demonstrating the model's efficiency for real-time applications. The k-fold cross-validation technique was also employed to ensure the model's generalization ability to new data. This study demonstrated that CNN is an effective method for classifying paprika varieties based on color analysis, offering an accurate, fast, and scalable solution for automating sorting and grading processes in the agricultural sector, reducing human errors, and improving operational efficiency.