These days I am working with a Jetson TX2 from Nvidia. It is a very powerful tiny computer, featuring a Tegra GPU with 256 cores. It allows me to run algorithms for mapping and visual localisation in real-time for PARA.
What you see is the development board, the card is actually only the part with the heat-spreader. It is the size of a credit card.
Here is my cmake command to compile OpenCV3:
cmake \
-DCMAKE_BUILD_TYPE=Release \
-DCMAKE_INSTALL_PREFIX=/usr \
-DBUILD_PNG=OFF \
-DBUILD_TIFF=OFF \
-DBUILD_TBB=OFF \
-DBUILD_JPEG=OFF \
-DBUILD_JASPER=OFF \
-DBUILD_ZLIB=OFF \
-DBUILD_EXAMPLES=OFF \
-DBUILD_opencv_java=OFF \
-DBUILD_opencv_python2=ON \
-DBUILD_opencv_python3=OFF \
-DENABLE_PRECOMPILED_HEADERS=OFF \
-DWITH_OPENCL=OFF \
-DWITH_OPENMP=OFF \
-DWITH_FFMPEG=ON \
-DWITH_GSTREAMER=ON \
-DWITH_GSTREAMER_0_10=OFF \
-DWITH_CUDA=ON \
-DENABLE_FAST_MATH=1 \
-DCUDA_FAST_MATH=1 \
-DWITH_CUBLAS=1 \
-DWITH_GTK=ON \
-DWITH_VTK=OFF \
-DWITH_TBB=ON \
-DWITH_1394=OFF \
-DWITH_OPENEXR=OFF \
-DCUDA_TOOLKIT_ROOT_DIR=/usr/local/cuda-9.0 \
-DCUDA_ARCH_BIN=6.2 \
-DCUDA_ARCH_PTX="" \
-DINSTALL_C_EXAMPLES=OFF \
-DINSTALL_TESTS=OFF \
-DBUILD_PERF_TESTS=OFF \
-DBUILD_TESTS=OFF \
../
And when the task is finished, run:
make -j4
sudo make install
It will compile fast because the tests and examples are OFF and the library will be CUDA enabled with the latest version from JetPack (9.0). I hope it will help someone out there :).