Comparison ADAM-optimizer and SGDM for Classification Images of Rice Leaf Disease

Abstract

One of the most important sectors related to the food security of a country is the agricultural sector. Considering rice plants are currently the largest commodity, the amount of rice production becomes the biggest concern. One of the factors influencing the fertility of rice is the rice plants diseases. The innovation are conducted to overcome this problem by collaborating with various sectors, one of them is computer science, especially computer vision. Deep learning methodology is a branch of computer vision and convolutional neural network (CNN). It is often used since it can compress features automatically and more efficiently. Hence, it can produce the accurate classification. Alexnet is an architecture or model on CNN for images classification. This study compared two optimizers, namely the Adam and SGDM optimizer. The result study obtained the best validation accuracy of types classification of rice leaf is the Adam optimizer as 10 epochs and a learning rate of 0.0001 as 95.33 %. The Adam optimizer generate a preferable validation accuracy than SGDM optimizer in a training dataset by the slightly epochs. However, a large number of epochs SGDM are better in training epochs.