PIGEON: Predicting Image Geolocations

arXiv:2307.05845v5 Announce Type: replace-cross
Abstract: Planet-scale image geolocalization remains a challenging problem due to the diversity of images originating from anywhere in the world. Although approaches based on vision transformers have made significant progress in geolocalization accuracy, success in prior literature is constrained to narrow distributions of images of landmarks, and performance has not generalized to unseen places. We present a new geolocalization system that combines semantic geocell creation, multi-task contrastive pretraining, and a novel loss function. Additionally, our work …

abstract accuracy arxiv cs.cv cs.lg diversity generalized image images literature narrow performance planet prior progress scale success transformers type vision vision transformers world