Dataset Card for BBBicycles

Dataset Summary

Bent & Broken Bicycles (BBBicycles) dataset is a benchmark set for the novel task of damaged object re-identification, which aims to identify the same object in multiple images even in the presence of breaks, deformations, and missing parts. You can find an interactive preview here.

Dataset Structure

The final dataset contains:

• Total of 39,200 image
• 2,800 unique IDs
• 20 models
• 140 IDs for each model

Information for each ID:	Information for each render:
Model Type Texture type Stickers	Background Viewing Side Focal Length Presence of dirt

Citation Information

@inproceedings{bbb_2022,
  title={Bent & Broken Bicycles: Leveraging synthetic data for damaged object re-identification},
  author={Luca Piano, Filippo Gabriele Pratticò, Alessandro Sebastian Russo, Lorenzo Lanari, Lia Morra, Fabrizio Lamberti},
  booktitle={2022 IEEE Winter Conference on Applications of Computer Vision (WACV)},
  year={2022},
  organization={IEEE}
}

Credits

The authors gratefully acknowledge the financial support of Reale Mutua Assicurazioni.