This is a course project for PolyU COMP 6815.
I only test this code on Win8.1 + VS2013.
- OpenCV 3.0 (or later)
- ArrayFire
Warning: This project is not well-optimized since it is only for understanding some detail operations in CNNs. I highly suggest that everyone should not use this code to train a large-scale network on large-scale dataset. Please use other excellent pacakges, such as Caffe, MxNet, MatConvnet, Cudnn etc.
- Install OpenCV3.0 and ArrayFie.
- Create a new Project in VS2013, and add all "CNN/*.h" and "CNN/.cpp" in "header" and "source" fold in VS2013.
- Make sure all "include" and "lib" in the right place. (You should include ArrayFire and OpenCV).
- Add "opencv_world300.lib" (or "opencv_world300d.lib") and "afcpu.lib" into "Additional Dependencies".
- Add "test/testMNIST" or "test/testCIFAR10" into "source" fold and run the project.
Now, this code can only run on CPU, so it is a little slower.