This dataset provides a platform to benchmark transfer-learning algorithms, in particular attribute base classification and zero-shot learning [1]. It can act as a drop-in replacement to the original Animals with Attributes (AwA) dataset [2,3], as it has the same class structure and almost the same characteristics.
It consists of 37322 images of 50 animals classes with pre-extracted feature representations for each image. The classes are aligned with Osherson's classical class/attribute matrix [3,4], thereby providing 85 numeric attribute values for each class. Using the shared attributes, it is possible to transfer information between different classes.
The image data was collected from public sources, such as Flickr, in 2016. In the process we made sure to only include images that are licensed for free use and redistribution, please see the archive for the individual license files. If the dataset contains an image for which you hold the copyright and that was not licensed freely, please contact us at , so we can remove it from the collection.


Please cite the following paper when using the dataset: Attribute based classification and the original Animals with Attributes (AwA) data is described in:
  • [2] C. H. Lampert, H. Nickisch, and S. Harmeling. "Learning To Detect Unseen Object Classes by Between-Class Attribute Transfer". In CVPR, 2009 (pdf)
  • [3] C. H. Lampert, H. Nickisch, and S. Harmeling. "Attribute-Based Classification for Zero-Shot Visual Object Categorization". IEEE T-PAMI, 2013 (pdf)
The class/attribute matrix was originally created by:
  • [4] D. N. Osherson, J. Stern, O. Wilkie, M. Stob, and E. E. Smith. "Default probability". Cognitive Science, 15(2), 1991.
  • [5] C. Kemp, J. B. Tenenbaum, T. L. Griffiths, T. Yamada, and N. Ueda. "Learning systems of concepts with an infinite relational model". In AAAI, 2006.


Version 1.0, June 9th 2017
The dataset consists of three archives:

  • Base package (32K) including the class/attribute table:AwA2-base.zip (same as for AwA1)
  • Features extracted using an ILSVRC-pretrained ResNet101 (as used in [1]): AwA2-features.zip
    • Note: The ILSVRC classes overlap partially with the AwA2 classes. Please use only the new proposed class-split (see [1]) when using pretrained CNN features!
  • Images in JPEG format and license files: AwA2-data.zip (13 GB!)
  • For the dataset split and pre-extracted features, please see here.
  • If you have problems downloading the newly proposed split file (v2.0) from there, here's a copy: xlsa17.zip

  • Results

    Zero-shot classification results (class-averaged multiclass accuracy [in %])

      (results from [1])
      AwA1 AwA2
      Method original split proposed split (preferred) original split proposed split (preferred)
      DAP [a] 57.1 44.1 58.7 46.1
      IAP [a] 48.1 35.9 46.9 35.9
      CONSE [b] 63.6 45.6 67.9 44.5
      CMT [c] 58.9 39.5 66.3 37.9
      SSE [d] 68.8 60.1 67.5 61.0
      LATEM [e] 74.8 55.1 68.7 55.8
      ALE [f] 78.6 59.9 80.3 62.5
      DEVISE [g] 72.9 54.2 68.6 59.7
      SJE [h] 76.7 65.6 69.5 61.9
      ESZSL [i] 74.7 58.2 75.6 58.6
      SYNC [j] 72.2 54.0 71.2 46.6
      SAE [k] 80.6 53.0 80.7 54.1
      GFZSL [l] 80.5 68.3 79.3 63.8
      • [a] C. Lampert, H. Nickisch, and S. Harmeling, “Attribute-based classification for zero-shot visual object categorization,” TPAMI, 2013.
      • [b] M. Norouzi, T. Mikolov, S. Bengio, Y. Singer, J. Shlens, A. Frome, G. Corrado, and J. Dean, “Zero-shot learning by convex combination of semantic embeddings,” ICLR, 2014.
      • [c] R. Socher, M. Ganjoo, C. D. Manning, and A. Ng, “Zero-shot learning through cross-modal transfer,” NIPS, 2013.
      • [d] Z. Zhang and V. Saligrama, “Zero-shot learning via semantic similarity embedding,” ICCV, 2015.
      • [e] Y. Xian, Z. Akata, G. Sharma, Q. Nguyen, M. Hein, and B. Schiele, “Latent embeddings for zero-shot classification,” CVPR, 2016
      • [f] Z. Akata, F. Perronnin, Z. Harchaoui, and C. Schmid, “Label-embedding for image classification,” TPAMI, 2016.
      • [g] A. Frome, G. S. Corrado, J. Shlens, S. Bengio, J. Dean, M. A. Ranzato, and T. Mikolov, “Devise: A deep visual-semantic embedding model,” NIPS, 2013
      • [h] Z. Akata, S. Reed, D. Walter, H. Lee, and B. Schiele, “Evaluation of output embeddings for fine-grained image classification,” CVPR, 2015.
      • [i] B. Romera-Paredes and P. H. Torr, “An embarrassingly simple approach to zero-shot learning,” ICML, 2015
      • [j] S. Changpinyo, W.-L. Chao, B. Gong, and F. Sha, “Synthesized classifiers for zero-shot learning,” CVPR, 2016
      • [k] E. Kodirov, T. Xiang, and S. Gong, “Semantic autoencoder for zero-shot learning,” CVPR, 2017.
      • [l] V. K. Verm and P. Rai, “A simple exponential family framework for zero-shot learning,” ECML, 2017.