Klasifikasi Citra Makanan Berdasarkan Asal Daerah Menggunakan Convolutional Neural Network

Indonesia's culinary culture has distinctive features and characteristics that vary from region to
region. Fast food and fast food are the favorite choices of young people today because they are
more practical and save time, while traditional food has decreased interest which can threaten
the preservation of the archipelago's culinary heritage. One of the objectives of this research is
to help preserve and promote Indonesia's culinary wealth with classification techniques based on
provincial regions among young people. Data is taken from google images using a bot that
simulates human behavior when retrieving links on google images. There are many deep learning
and machine learning methods that can be used to classify food images, an example is CNN. The
results of using the CNN method show an accuracy value of 64% in predicting food images based
on regional origin. This result shows that there are several obstacles that need to be considered.
One of the reasons for this low accuracy is the complex data variation in food images from both
islands and certain visual similarities that are difficult to identify by the model, causing false
positives and false negatives. However, the CNN method is relatively good enough to be applied
to the classification of typical food images of Java Island and Sumatra Island.