Finding high-probability mobile photography routes between origin and destination endpoints

Abstract

A challenge for mobile smartphone photographers is finding the best route to take interesting photos given origin and destination constraints. To solve this problem, we leveraged crowd-sourced geotagged photographs in order to build a model that can generate the highest-probability route between two geolocation endpoints. Our implemented system comprises discretizing the geocoordinate space, determining popular geolocations, computing Markov transition probabilities, and finally generating highest-probability end-to-end routes by solving a shortest path problem where edge probabilities are converted into edge distances. The experimental results show the impact of path length and spatial resolution on the prediction accuracy of our system.