CNN Implementation in C++

This project implements a Convolutional Neural Network (CNN) from scratch in C++ for MNIST and CIFAR10 data classification using OpenBLAS and OpenMP.

Build Instructions

Be sure OpenBLAS is installed in OpenBLAS_DIR directory in CMakeLists.txt

At project root directory:

# Create build directory
mkdir -p out/build

# Move to build directory
cd out/build

# Generate build files
cmake ../..

# Build the project
cmake --build .

Running the Project

After building, while in the out/build directory:

Training

./CppCNN/mnist_train
./CppCNN/cifar10_train

Inference

./CppCNN/mnist_inference

Project Structure

.
├── CMakeLists.txt
├── CMakePresets.json
├── CppCNN/
│   ├── CMakeLists.txt
│   ├── data/               # MNIST dataset and model files
│   ├── include/            # Header files
│   └── src/               # Source files
├── LICENSE
└── README.md

Prerequisites

• C++20 compatible compiler
• CMake 3.0 or higher

Data

The project expects MNIST and CIFAR10 dataset in the CppCNN/data/mnist and CppCNN/data/cifar10 directory: In mnist directory:

• train-images.idx3-ubyte
• train-labels.idx1-ubyte
• test-images.idx3-ubyte
• test-labels.idx1-ubyte

In cifar10 directory:

• data_batch_N.bin
• test_batch.bin
• batches.meta.txt

After training, model files will be saved in the same directory with names like mnist_model_epoch_N.bin where N is the epoch number.