This dataset is strictly for academic use. The distributors of this dataset are not liable for any unintended uses of this dataset. This dataset is not to be redistributed without consent from the authors.

This dataset contains pairs of caricature images (in directory 'Caricature') and photograph images (in directory 'Photo'). There are 207 subjects, and each image is named as image_XXX.jpg, where XXX specifies the subject number.

Any use of this dataset in publications must properly cite the reference below. We prefer the any accuracy comparisons to be made using the true accept rates (TAR) at fixed false accept rates (FAR) of 1.0% and 10.0%. The reason is that these measures are not contingent on the number of subjects used in training and testing (unlike CMC/Rank scores).

Please note that the paper below used 196 pairs of caricatures and photos, while this set contains 207 pairs. This is the result of a data scrub that corrected 3 erroneous pairs (which in turn decreased the accuracy of the reported results), and the addition of 14 caricature/photo pairs.

Finally, please check back with the authors as a larger dataset will be released soon. 

Best of luck!

B.F. Klare, S. S. Bucak, T. Akgul, and A. K. Jain, "Towards Automated Caricature Recognition", in Proceedings of the 5th IAPR International Conference on Biometrics (ICB), New Delhi, India, March 29-April 1, 2012. 
http://www.cse.msu.edu/rgroups/biometrics/Publications/Face/Klareetal_TowardsAutomatedCaricatureRecognition_ICB12.pdf