Border Legibility with Computer Vision
Research Mentor(s)
Scott Wehrwein
Description
The physical manifestation of international political borders in overhead imagery presents a unique opportunity to study the effects of border-related policies on the real world. However, manual analysis of such imagery is labor-intensive and difficult to achieve at global scale. This project aims to address this by using computer vision techniques to automate analysis of overhead imagery of borders. The contributions of this ongoing work are threefold: (1) we propose the novel task of predicting the legibility of a border from an overhead image; (2) we collected the Borders-Overhead dataset, consisting of over 600,000 aerial images depicting borders between over 280 pairs of countries. (3) we leverage both classical and machine-learning based Computer Vision techniques to establish baselines for the border legibility task. We directly compute the success of our methods via human-annotated pairwise comparisons on 1,000 sample images from our dataset. Our legibility analysis constitutes one of the largest and most comprehensive visual analysis of political borders thus far and furthers progress towards a more comprehensive understanding of how political borders and border policies manifest in the physical world.
Document Type
Event
Start Date
May 2022
End Date
May 2022
Location
Carver Gym (Bellingham, Wash.)
Department
CSE - Computer Science
Genre/Form
student projects; posters
Type
Image
Rights
Copying of this document in whole or in part is allowable only for scholarly purposes. It is understood, however, that any copying or publication of this document for commercial purposes, or for financial gain, shall not be allowed without the author’s written permission.
Language
English
Format
application/pdf
Border Legibility with Computer Vision
Carver Gym (Bellingham, Wash.)
The physical manifestation of international political borders in overhead imagery presents a unique opportunity to study the effects of border-related policies on the real world. However, manual analysis of such imagery is labor-intensive and difficult to achieve at global scale. This project aims to address this by using computer vision techniques to automate analysis of overhead imagery of borders. The contributions of this ongoing work are threefold: (1) we propose the novel task of predicting the legibility of a border from an overhead image; (2) we collected the Borders-Overhead dataset, consisting of over 600,000 aerial images depicting borders between over 280 pairs of countries. (3) we leverage both classical and machine-learning based Computer Vision techniques to establish baselines for the border legibility task. We directly compute the success of our methods via human-annotated pairwise comparisons on 1,000 sample images from our dataset. Our legibility analysis constitutes one of the largest and most comprehensive visual analysis of political borders thus far and furthers progress towards a more comprehensive understanding of how political borders and border policies manifest in the physical world.