Key points are not available for this paper at this time.
Today, people increasingly absorb information about places from online images of locations they have never been to. In the context of contested territories, online images serve as a medium for imagining territories. However, online images of contested territories, just like any other type of mediated information, are channeled through political narratives and ideologies. To quantitatively characterize this digitally mediated territorial imagination, this article proposes a new deep learning approach using Google Cloud Vision API to quantitatively characterize scraped online images of contested territories queried through different toponyms. Through a case study of three disputed territories employing six toponyms, we find that the results of online image queries are different depending on toponymic inputs, and we show how this can lead to different territorial imaginations localized to political and historical contexts. This work contributes these findings using a novel approach that integrates quantitative analytics and critical perspectives on territorial imagination, highlighting future opportunities for deep learning methods to provide insights into real-world geopolitical issues. • Google Cloud Vision API's deep learning approach can analyze characteristics of online images of contested territories. • Different types of online images are retrieved with different toponyms of contested territories. • Variations in the retrieved images reflect various contextualized political narratives surrounding the contested territories.
Sui et al. (Sat,) studied this question.