This study aimed to determine the usefulness of different parts of the plum tree (leaf and flower) and fruit (skin and endocarp) to discriminate cultivars ‘Cacanska Lepotica’, ‘Kalipso’, ‘Opal’, ‘Polinka’, and ‘Węgierka Dąbrowicka’. The distinguishing models were developed based on features extracted from color images acquired using a flatbed scanner. The texture features from the images of plum endocarps proved to be the most useful in developing the most accurate models. The average accuracy of distinguishing plum endocarps ‘Cacanska Lepotica’, ‘Kalipso’, ‘Opal’, ‘Polinka’, and ‘Węgierka Dąbrowicka’ reached 94.8% for the machine learning model built using the Quadratic SVM algorithm. A slightly lower accuracy of 92.8% was obtained for the model developed based on plum leaf textures using the Wide Neural Network. In the case of plum flowers, the model built using the Quadratic SVM achieved an average accuracy of 90.4%. The lowest accuracies were obtained by CNN (convolutional neural network) models using color images. The obtained results revealed that the cultivar classification accuracy depended on the part of the plum used to build models. The results can be used in practice to select the part of the plum tree or fruit, which is the most suitable for distinguishing cultivars with the highest accuracy.
Ropelewska et al. (Fri,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: