Libtorch_install

Libtorch

Libtorch is C++ API of torch machine learning library.

Install

Download and unzip libtorch.
In CMakeLists.txt file of ExtendedObjectDetection project change line 12:

set(torch OFF)

to

set(torch ON)

And specify path to unzipped libtorch in line 43:

set(Torch_DIR "/home/${USER}/Lib/libtorch/share/cmake/Torch/")

Then rebuild project.

Install libtorch with CUDA on Jetson

From here https://discuss.pytorch.org/t/how-does-one-use-libtorch-c-with-the-arch64-build-of-pytorch-for-the-nvidia-jetson/140627/5

Firstly, you should download PyTorch for Jetson on below link

https://forums.developer.nvidia.com/t/pytorch-for-jetson/72048

Secondly, install the “whl” file following this guide

https://docs.nvidia.com/deeplearning/frameworks/install-pytorch-jetson-platform/index.html

After you finished the above steps. The LibTorch has been installed on your device.

How to Use?

where is LibTorch installed?

Commonly, it installs at ~/.local/lib/python3.6/site-packages/torch/ In this dir, you can see bin, lib, and include dir, It is organized just like other C++ libraries. How to use LibTorch with CMake?

for example, add these two lines to your CMakeLists.txt

// set PyTorch path
set(Torch_DIR /home/nx/.local/lib/python3.6/site-packages/torch/share/cmake/Torch)
// use find_package to use libtorch
find_package(Torch REQUIRED)