Paper Title
Automatic Restaurant Image Tagging
Abstract
Image tagging provides crucial information in many situations. This paper provides investigation of deep learning
techniques for this task. We applied these techniques on images taken from restaurant scenes where people tag information
only on a few of them. Our aim is to learn from the tagged images and then automatically tag the rest of images. We
investigated two deep learning techniques: AlexNet and GoogLeNet. The two techniques are measured first without any
pre-trained and then pre-trained with ILSRVC-12 and Places365 dataset. The result of these two techniques provided relatively
high tagging accuracy even without pre-train. However, with pre-trained model, the accuracy risen significantly.
Index Terms—Deep Learning, Image Classification, Image Tagging,