Tourists’ perceptions of urban space: a computer vision approach

Abstract

Purpose This study aims to conduct an empirical investigation of differing perceptions of nine types of urban space and nine visual elements among tourists in destination using a computer vision (CV) approach. Design/methodology/approach The data for this study was extracted from YFCC 100 M dataset. Nine types of urban space in Beijing were initially identified using a scene recognition model. Subsequently, a semantic segmentation model was applied, which yielded substantial evidence relating to nine visual elements that were used to elicit differing perceptions among tourists from different continents. Findings Tourists from three continents had different perceptions about corridors, old buildings, overlooks and traffic spaces, reflecting their cultural convention. Asians, Europeans and North Americans diversely gazed at the landscape element of buildings, foliage, sky and people in urban space. All those provided evidence to contribute to the tourist gaze theory's construction. Originality/value This study firstly depicted how tourists perceive the tourism symbol of urban space. The novel approach of employing two CV models offer methodological insights to tourism research relevant to visual perception.

Camera eats first: exploring food aesthetics portrayed on social media using deep learning

“…It may directly assign aesthetic scores to food images that reflect the image quality without further processing. However, existing hospitality literature adopt deep learning methods to perform basics tasks such as image classification, scene and entity recognition (Cho et al , 2022; Zhang et al , 2021b,c). On the other hand, we use computer vision methods to obtain a large set of low-level visual features that are indicative of aesthetic qualities from the computer science field (Datta et al , 2006; Deng et al , 2017; Machajdik and Hanbury, 2010; Wang et al , 2013).…”

“…We provide a list of tourism and hospitality research papers using visual data analytics on social media in Table 1, reporting research gaps, image data and the proposed solutions, respectively. Given its great potential in extracting the complex structural representation of images at large scale, instead of labor-intensive manual encodings that work with hundreds of images (Stepchenkova and Zhan, 2013), these studies leverage computer vision and deep learning techniques to assess cultural ecosystems service (Richards and Tunçer, 2018), recommend advertising images for tourism destinations (Deng and Li, 2018), recognize the entities appearing in images (Ren et al , 2021; Wang et al , 2020), investigate tourists’ behavior and perception (Zhang et al , 2019a, 2021b,c) and explore tourists’ photos reflecting regional characteristics (Cho et al , 2022). However, none of these studies provides the image quality assessment, which is important in visual data analytics on social media.…”