Tourist Mobility Patterns: Faster R-CNN Versus YOLOv7 for Places of Interest Detection

Abstract

The mobility of tourists plays a significant role in shaping their travel experiences and the overall dynamics of a destination. In recent years, the proliferation of social media platforms has provided a rich source of visual data, allowing us to leverage the abundance of pictures shared by tourists to extract meaningful information. Using computer vision techniques and deep learning algorithms, such as object detection, it becomes possible to extract useful information from tourist pictures. In this study, we look for the best way to detect objects from pictures shared by tourists during their journey in order to determine their locations. To achieve our goal we propose a new methodology composed by; database creation, database annotation, preprocessing, deep learning implementation and evaluation. We implemented two deep learning object detection methods: YOLOv7 and Faster R-CNN. A dataset has been created to provide examples of training and testing for neuronal networks. The training was performed on various basic models, in order to increase the efficiency of the training time and to compare the results. We evaluated the results using three parameters: precision, recall and mAP. The results indicate that YOLOv7 has the precision and performance, with over 90 % mAP, 92.1 % precision and 92.7 % recall.