C++ implementation of neural networks library with Keras-like API. Contains majority of commonly used layers, losses and optimizers. Supports sequential and multi-input-output (flow) models. Supports single CPU, Multi-CPU and GPU tensor operations (using cuDNN and cuBLAS).

Arbitrary style transfer to content Images through "Adaptive Instance Normalization" (Ada-IN) and "Whitening and coloring transforms" (WCT)