Final project for ECE 408 at UIUC
It's a CUDA-accelerated image recognition training program. Specifically, given a set of input images and corresponding labels, it uses CUDA to find all the significant Haar-like features in all the images, then uses AdaBoost to identify the features that most accurately describe the object being trained.
Open VSProject/408Final.v12.suo in Visual Studio 2013 and run. CUDA 7.5 doesn't seem to support VS2015 yet. It uses images from the MNIST handwritten digit sample set to create a strong classifier for the digit '0'. Note that this was done for an ECE408 project, and the actual identifier for the created strong classifier was not part of the project scope. As a result, you can't actually use it to identify digits.