Scaled and Translated Image Recognition (STIR) Source Data

Thomas Altstidl, An Nguyen, Leo Schwinn, Franz Köferl, Christopher Mutschler, Bjoern M. Eskofier, Dario ZancaZenodo 2022 · DOI · OpenAlex

Abstract

While convolutions are known to be invariant to (discrete) translations, scaling continues to be a challenge and most image recognition networks are not invariant to them. To explore these effects, we have created the Scaled and Translated Image Recognition (STIR) dataset. This dataset contains objects of size \(s \in [17,64]\), each randomly placed in a \(64 \times 64\) pixel image. <strong>Original Source Data</strong> <code>dota/</code> (from DOTA v1.5 Google Drive website) <code>train/</code> <code>DOTA-v1.5_train.zip</code> <strong>not</strong> unzipped <code>part1.zip</code> <strong>not</strong> unzipped <code>part2.zip</code> <strong>not</strong> unzipped <code>part3.zip</code> <strong>not</strong> unzipped <code>val/</code> <code>DOTA-v1.5_val.zip</code> <strong>not</strong> unzipped <code>part1.zip</code> <strong>not</strong> unzipped <code>fontawesome/</code> (from Font Awesome 5.15.3 "Free for Desktop") <code>svgs/</code> unzipped from archive <code>mapillary/</code> (from Mapillary Traffic Sign Dataset) <code>mtsd_v2_fully_annotated</code> unzipped from archive <code>train.0.zip</code> <strong>not</strong> unzipped <code>train.1.zip</code> <strong>not</strong> unzipped <code>train.2.zip</code> <strong>not</strong> unzipped <code>val.zip</code> <strong>not</strong> unzipped <code>mnist/</code> (from Yann LeCun website) <code>t10k-images-idx3-ubyte.gz</code> <code>t10k-labels-idx1-ubyte.gz</code> <code>train-images-idx3-ubyte.gz</code> <code>train-labels-idx1-ubyte.gz</code> <strong>License and Attribution</strong> When using the original source data for your own research, please respect the individual licenses. For attribution in papers, we recommend the following citations which introduce the respective datasets. D. Gandy, J. Otero, E. Emanuel, F. Botsford, J. Lundien, K. Jackson, M. Wilkerson, R. Madole, J. Raphael, T. Chase, G. Taglialatela, B. Talbot, and T. Chase. Font Awesome. https://fontawesome.com/v5/download, Nov. 2022. Y. Lecun, L. Bottou, Y. Bengio, and P. Haffner. Gradient-based learning applied to document recognition. <em>Proc. IEEE</em>, 86(11):2278–2324, Nov. 1998. C. Ertler, J. Mislej, T. Ollmann, L. Porzi, G. Neuhold, and Y. Kuang. The Mapillary Traffic Sign Dataset for Detection and Classification on a Global Scale. In <em>2020 16th Eur. Conf. Comput. Vision (ECCV)</em>, Glasgow, UK, Aug. 2020. G.-S. Xia, X. Bai, J. Ding, Z. Zhu, S. Belongie, J. Luo, M. Datcu, M. Pelillo, and L. Zhang. DOTA: A Large-Scale Dataset for Object Detection in Aerial Images. In <em>2018 IEEE/CVF Conf. Comput. Vision and Pattern Recognition (CVPR)</em>, pages 3974–3983, Salt Lake City, UT, USA, June 2018.
BibTeX
@inproceedings{altstidl2022scaled,
  author = {Thomas Altstidl and An Nguyen and Leo Schwinn and Franz Köferl and Christopher Mutschler and Bjoern M. Eskofier and Dario Zanca and Bjoern Eskofier},
  title = {{Scaled and Translated Image Recognition (STIR) Source Data}},
  booktitle = {Zenodo},
  year = {2022},
  doi = {10.5281/zenodo.7351725},
}

Loading related content...

Figure