VIPPGeo Dataset

This is the official repository for the paper :

Which country is this picture from? New data and methods for DNN-based country recognition

Download the dataset urls from google drive

Or

By using gdown

pip install gdown
gdown https://drive.google.com/drive/folders/1CXVdpfFpolQah4PsfGXrhgoWtOtFEC__ --folder

After downloading the urls, download the dataset, using the follwoing scripts. To store the datast in msg fromt (to save space), please specify --msg option.

Get your Access token for Mapillary API https://www.mapillary.com/developer/api-documentation  

Train

python3 download_images.py --output dir_out/train --url_csv GeoDataset_Urls/train.csv --access_token_Mapillary  MLY|################|********************************

Validation

python3 download_images.py --output dir_out/validation --url_csv GeoDataset_Urls/valid.csv  --access_token_Mapillary  MLY|################|********************************

Test

python3 download_images.py --output dir_out/test --url_csv GeoDataset_Urls/test.csv --access_token_Mapillary  MLY|################|********************************

In addition, you can download our test set (the images and the necessary metadata) from google drive.

pip install gdown
gdown https://drive.google.com/drive/folders/1En2vYbd02J2RGzOICOsMA2N1UuHJhsx4 --folder

Citation

If you use this code for your research, please cite our paper.

@article{alamayreh2022country,
  title={Which country is this picture from? New data and methods for DNN-based country recognition},
  author={Alamayreh, Omran and Dimitri, Giovanna Maria and Wang, Jun and Tondi, Benedetta and Barni, Mauro},
  journal={arXiv preprint arXiv:2209.02429},
  year={2022}
}

Acknowledgement

This code heavily depends on the code provided by

Geolocation Estimation of Photos using a Hierarchical Model and Scene Classification