Betawi Traditional Food Image Detection using ResNet and DenseNet

Abstract

Technological developments in the field of Smart System is now growing and began to spread to various areas such as tourism sector. In this research, we developed a smart system for Betawi culinary tourism. Detection of traditional food names using images is a challenge because the variety of shape and direction of shooting is always different. The use of deep learning architecture is expected to overcome the problem, but the selection of effective deep learning architecture is also a problem. This study compares some deep learning architecture to determine the suitable architecture to detect culinary images. Based on our experimental results, DenseNet169 gives the best performance in terms of accuracy, error rate and training time when using CPU and ResNet50 when using GPU..