Automatic evaluation of image processing algorithms.

We want to automatically evaluate the correctness of image processing code in the context of online courses. Steudents will submit python code, which we will run on new input images. The main interesting challenge is that multiple answers might be "correct" from the perspective of the course and we need to develop comparison tools that take into account this set of possibilities.

Contact: Fredo Durand,fredo@csail.mit.edu