Tomato Maturity Estimation Using Deep Neural Network

In this study, we propose a tomato maturity estimation approach based on deep neural network. Tomato images were obtained using an RGB camera installed monitoring robot and samples cropped to generate dataset with which train the classification model. The model is trained cross-entropy loss mean–variance loss, can implicitly provide label distribution knowledge. For continuous in test stage, output probability of four classes calculated as expected (normalized) value. Our results demonstrate that F1 score was approximately 0.91 average, range 0.85–0.97. Furthermore, comparison hue value—which correlated growth—showed no significant differences between estimated values, except pink stage. From overall results, found our not only classify discrete maturation stages tomatoes but also continuously estimate their maturity. it higher accuracy data labeling, more precise may be achieved.