README.md

metadata

task_categories:
  - image-classification
  - image-to-text

Dataset Card for MP16-Pro Dataset

This dataset contains 4,654,532 geo-tagged images sourced from Flickr. As time has passed, some image links have become invalid, and the number of images currently available is approximately 4.12 million. This version of dataset is compiled in January 2024.

Dataset Details

Dataset Sources

• Repository: https://github.com/Applied-Machine-Learning-Lab/G3
• Paper: G3: An Effective and Adaptive Framework for Worldwide Geolocalization Using Large Multi-Modality Models
• Meta Data: GeoEstimation
  • url data: metadata/mp16-urls.csv
  • meta data: metadata/MP16_Pro_places365.csv
  • filtered meta data: metadata/MP16_Pro_filtered.csv (invalid images are filtered out)

Uses

Direct Use

cat mp-16-images* > mp-16-images.tar

Dataset Creation

Source Data

• mp16-urls
• mp16-places365

Data Collection and Processing

Images are downloaded based on mp16-urls. The / in all image names has been replaced with _.

Who are the source data producers?

Flickr

Annotations

Bias, Risks, and Limitations

Please adhere to the requirements of the MP16 dataset and use it solely for academic research purposes.

Citation

BibTeX:

@article{jia2024g3,
  title={G3: an effective and adaptive framework for worldwide geolocalization using large multi-modality models},
  author={Jia, Pengyue and Liu, Yiding and Li, Xiaopeng and Zhao, Xiangyu and Wang, Yuhao and Du, Yantong and Han, Xiao and Wei, Xuetao and Wang, Shuaiqiang and Yin, Dawei},
  journal={Advances in Neural Information Processing Systems},
  volume={37},
  pages={53198--53221},
  year={2024}
}