The proposed project is an examination of different methods for image style transfer, from a linear algebraic construction through automatic and convolutional neural networks. Hopefully to understand what patterns and features a machine can learn that underlie human image perception.
With a working installation of Julia, handily provided via the JuliaPro project for ease of installation.
Update any packages that need updating, Pkg.update() at the REPL. This project requires adding the following Julia packages:
- Images.jl
- ImageSegmentation
- Colors.jl
- Cxx.jl (requires cmake 3.4.3+)
- MATLAB.jl
We're going to use the dlib C++ library to do facial landmark detection on our input and reference images. We won't need the entire library, because we're going to bootstrap the landmark detection code from Julia.
- JuliaPro Project
- dlib