Classification of Indonesian Coffee Types with Deep Learning
| Main Authors: | Rivalto, Alfan, Pranowo, ., Santoso, Albertus Joko |
|---|---|
| Format: | BookSection PeerReviewed Book |
| Bahasa: | eng |
| Terbitan: |
AIP Publishing
, 2020
|
| Subjects: | |
| Online Access: |
http://e-journal.uajy.ac.id/26665/1/33.%20Classification%20of%20Indonesian%20coffee%20types%20with%20deep%20learning.pdf http://e-journal.uajy.ac.id/26665/ |
Daftar Isi:
- Indonesia is one of the largest coffee producing and exporting countries in the world. The development of the coffee business has progressed quite rapidly, starting from the level of farmers, suppliers, coffee cafes, to ordinary consumers. Besides the increasing progress of the coffee industry in Indonesia, there are still many problems that cause material losses and a sense of dissatisfaction for both business and coffee lovers. The problem that arises is because the industry is still run a lot by using a system of trust between the parties concerned. It is difficult for a simple system to distinguish between one coffee variant and another. The need for an information technology-based system that can help identify and ensure directly that the coffee needed and enjoyed is in accordance with what is desired. The information system that will be built can classify the types of coffee based on the image. The introduction of these image patterns uses Deep Learning. Training the Deep Learning algorithm to detect coffee types accurately requires a large number of images for training data. The recognition method uses the convolutional neural network which can be used to recognize objects in an image and is often used to classify data in the form of images. The current CNN method trend is used for image classification problems due to the very high level of accuracy. CNN will classify each image prepared as training data for the introduction. Data is collected by taking pictures of coffee beans using a camera. This data collection contains 4 types of coffee from Indonesia (Garut, Gayo, Kerinci, Temanggung) with 617 images of coffee beans. After testing, the system can recognize objects with an accuracy of 74.26%.