Don Geman and I developed a series of synthetic image recognition problems which require to go beyond local descriptors and simple statistics of the image to be solved properly, and that we used to compare humans and machine-learning (Fleuret et al. 2011b).
This software should compile on any standard GNU/Linux machine, and UNIX machine in general. It only requires the jpeg and png library.
To test it on a list of problems, just run the script doit.sh with the problem numbers as arguments. For instance:
./doit.sh 1 2 3It will compile the source and run the executable on the specified problems to generate sample images and test a Boosted classifier. If no argument is given, it will run on all the problems by default.