Multi-Label Classification of Jasmine Rice Germination Using Deep Neural Network

Abstract

This paper proposes a multi-label image classification of Jasmine Rice (Thai Hom Mali) seed germination using the Deep Neural Network architecture. First, we have collected the dataset of normal germination of the rice and separated them into three classes: excellent-germination, good-germination, and poor-germination. Second, we feed the dataset into the Convolutional Neural Network for multi-label classifications. The dataset consists of 970 pictures in the training set and 194 images in the validation set. Lastly, we evaluated the model based on the confusion matrix. The results show that the precision, recall, and Fl-score are 0.80, 1.00, and 0.89 for excellent germination, 0.83, 0.83, and 0.83 for good germination, and 1.00, 0.87, 0.93 for poor germination, respectively. The accuracy of the predictions is satisfactory, which is higher than 0.89.