Open Set Logo Detection Dataset (OSLD Dataset, formerly known as BLAC dataset) is a dataset of eCommerce product images with associated brand logo images (see paper for details). It is released under creative commons (CC BY-NC 4.0) license to promote research in open set logo detection. The dataset can be used only for research purposes. It contains:

• 20K product images, with logo bounding box annotations (this dataset loader in PyTorch can be useful to load and evaluate)
• 12.1K logo classes with 20.8K canonical logo images
• Product image logo bounding box to canonical logo image match pair annotations

Dropbox

• 30 August 2021: Initial release of the dataset

For questions, please contact Muhammet Bastan ([email protected])

@article{OSLD,
  author    = {Muhammet Bastan and
               Hao-Yu Wu and
               Tian Cao and
               Bhargava Kota and
               Mehmet Tek},
  title     = {{Large Scale Open-Set Deep Logo Detection}},
  year      = {2019},
  url       = {http:https://arxiv.org/abs/1911.07440}
}