GeoDE: a Geographically Diverse Evaluation Dataset for Object Recognition

TL;DR

GeoDE is a new geographically diverse image dataset designed to reduce biases and PII issues in object recognition, collected via global solicitation to improve model evaluation and training.

Contribution

The paper introduces GeoDE, a novel dataset with global geographic diversity, collected through direct solicitation, addressing biases and privacy concerns in existing datasets.

Findings

GeoDE reduces geographic bias in object recognition datasets.

Models trained on GeoDE show improved generalization across regions.

GeoDE highlights limitations of current models in diverse settings.

Abstract

Current dataset collection methods typically scrape large amounts of data from the web. While this technique is extremely scalable, data collected in this way tends to reinforce stereotypical biases, can contain personally identifiable information, and typically originates from Europe and North America. In this work, we rethink the dataset collection paradigm and introduce GeoDE, a geographically diverse dataset with 61,940 images from 40 classes and 6 world regions, with no personally identifiable information, collected by soliciting images from people around the world. We analyse GeoDE to understand differences in images collected in this manner compared to web-scraping. We demonstrate its use as both an evaluation and training dataset, allowing us to highlight and begin to mitigate the shortcomings in current models, despite GeoDE's relatively small size. We release the full dataset and code at https://geodiverse-data-collection.cs.princeton.edu