Classification of Tomato Fruit Ripeness Level Using Convolutional Neural Network–Support Vector Machine Based on Digital Image

Tomato ripeness classification is an important task in post-harvest quality management, as the ripeness level directly influences taste, shelf life, and market value. Conventional ripeness assessment methods based on manual visual inspection are inherently subjective and often yield inconsistent results. To address this limitation, this study proposes an image-based tomato ripeness classification model using a hybrid Convolutional Neural Network–Support Vector Machine (CNN–SVM) approach. In the proposed model, a pretrained ResNet-50 architecture is employed as a fixed feature extractor to derive deep visual representations, while a Support Vector Machine with a Radial Basis Function kernel is utilized for final classification. The model is evaluated using a publicly available tomato image dataset, with the analysis limited to unripe and ripe categories. Image preprocessing procedures include resizing, normalization, and data augmentation, followed by an 80:20 train–test split strategy. Experimental results demonstrate that the proposed CNN–SVM model achieves strong and balanced performance, with an accuracy of 96.56%, a weighted precision of 96.80%, a recall of 96.56%, and an F1-score of 96.57%. These findings indicate that integrating deep feature extraction with an SVM classifier provides an effective and robust solution for tomato ripeness classification, particularly under limited data conditions.