KLASIFIKASI JENIS BUAH KURMA MENGGUNAKAN METODE CONVOLUTION NEURAL NETWORK

Dates are widely recognized by the people of Indonesia, dates are widely produced
in the middle eastern country of KSA and dates have various types so that it is
difficult for people to distinguish each type of date fruit. Therefore, it is necessary
to create a system to detect the types of dates according to their type. This research
focuses on classifying the types of dates. The dataset used in this study is 9 types of
date fruit collected with a total of 1658 date fruits. The author uses transfer learning
on a pre-trained model and uses a Convolution Neural Network (CNN) with
different components from previous research to extract features on all images in
the database. The extraction features used consist of 12 layers from the CNN model,
the difference in the first model using the Adam optimizer with a learning rate of
0.0001 epoch 20 & 40 trials which get a maximum value of 96% and the second
model with a learning rate of 0.0001 optimizer RMSprop with the same epoch two
trials get more maximum results, namely 100%. This study proves the suitability of
the database with a learning rate of 0.0001 and the RMSprop optimizer on this
CNN model with results that match the objectives of this study. For further
research, the application of other models such as Resnet or others can be used in
date fruit database experiments.