This is an official repository of the IMDB-WIKI-SbS dataset presented in this paper.

└── data
    ├── crowd_labels.csv
    └── gt.csv

The crowd_labels.csv file contains the annotation. It is a comma-separated table containing information about the comparisons made by crowd performers:

• left. URL of the image positioned on the left side of the task's page.
• right. URL of the image positioned on the right side of the task's page.
• label. URL of the image of the older person according to the annotator's opinion.
• performer. Anonymized annotator's identifier.

The gt.csv file contatins the ground-truth information from the original IMDB-WIKI dataset. It is a comma-separated table with columns:

• label. Image url.
• score. The ground-truth age in years.

Please cite the challenge results or dataset description as follows.

• Ustalov D., Pavlichenko N. IMDB-WIKI-SbS: An Evaluation Dataset for Crowdsourced Pairwise Comparisons.

@article{DBLP:journals/corr/abs-2110-14990,
  author    = {Nikita Pavlichenko and
               Dmitry Ustalov},
  title     = {IMDB-WIKI-SbS: An Evaluation Dataset for Crowdsourced Pairwise Comparisons},
  journal   = {CoRR},
  volume    = {abs/2110.14990},
  year      = {2021},
  url       = {https://arxiv.org/abs/2110.14990},
  eprinttype = {arXiv},
  eprint    = {2110.14990},
  timestamp = {Tue, 02 Nov 2021 15:31:04 +0100},
  biburl    = {https://dblp.org/rec/journals/corr/abs-2110-14990.bib},
  bibsource = {dblp computer science bibliography, https://dblp.org}
}