mini-dnn-cpp
                                
                                
                                
                                    mini-dnn-cpp copied to clipboard
                            
                            
                            
                        C++ demo of deep neural networks (MLP, CNN)
mini-dnn-cpp
mini-dnn-cpp is a C++ demo of deep neural networks. It is implemented purely in C++, whose only dependency, Eigen, is header-only.
Usage
Download and unzip MNIST dataset in mini-dnn-cpp/data/mnist/.
mkdir build
cd build
cmake ..
make
Run ./demo.
Result:
simple neural network with 3 FC layers can obtain 0.97+ accuracy on MNIST testset.
LeNet-like CNN can obtain 0.98+ accuracy on MNIST testset.