DTT is a header-only library that provides data transfer tools between C++ linear algebra libraries. Currently, it supports data transfer between the following libraries:

• Eigen (>=3.3.x)
• Armadillo (>=9.x)
• OpenCV (cv::Mat) (2.x, 3.x, 4.x)
• ArrayFire (>=3.x)
• LibTorch (PyTorch C++) (>=1.x)

Current status:

Last page update: 25/08/2019

Tested on:

• MacBook Pro (13-inch, 2017)
• Mac OS X Mojave (10.14.5), Clang 10 (clang-1000.11.45.5), Xcode 10.1 with the following libraries:
• • OpenCV 4.1.1 (stable, built from source)
• • Eigen 3.3.7 (stable)
• • Armadillo 9.600.5 (stable)
• • ArrayFire 3.6.4 (stable)
• • LibTorch (1.3.0.dev20190820)

Install dependencies:

brew install eigen
brew install armadillo
# download and install ArrayFire: https://arrayfire.com/download/
# download and install LibTorch: https://pytorch.org/get-started/locally/
# download adn install OpenCV: https://opencv.org/releases/

How to compile and run:

git clone https://github.com/andrewssobral/dtt.git
cd dtt && mkdir build && cd build
cmake -DCMAKE_PREFIX_PATH=$LIBTORCH_HOME ..
./dtt_test

How to use:

#include <dtt.h>
using namespace dtt;
// that's all!

List of available functions:

• From Eigen:

• • eigen2arma
• • eigen2cv
• • eigen2af
• • eigen2libtorch

• From Armadillo:

• • arma2eigen
• • arma2cv
• • arma2af
• • arma2libtorch

• From OpenCV:

• • cv2eigen
• • cv2arma
• • cv2af
• • cv2libtorch

• From ArrayFire:

• • af2eigen
• • af2arma
• • af2cv
• • af2libtorch

• From LibTorch:

• • libtorch2eigen
• • libtorch2arma
• • libtorch2cv
• • libtorch2af

See test/dtt_test.h and test/dtt_test.cpp for usage examples.