ThermalORB

Table of Contents

1. General Info
2. Installation
3. Running Instructions

General Info

---

C++ Wrapper of https://github.com/eric-yyjau/pytorch-superpoint using PyTorch and OpenCV. The code is flexible to different SuperPoint weights because it take a Pytorch model file (.pt) as input.

Installation

---

Requires

• C++14
• PyTorch >= 1.1 (tested in 1.7.1)
• OpenCV >= 4 (will be extended to 3 soon)
• [OPTIONAL] CUDA >= 10

$ git clone https://github.com/mattsara24/ThermalORB
$ cd ../path/to/the/file
$ mkdir build && cd build
$ cmake ..
$ make

NOTE

In order to properly compile the code you must edit the CMakeLists.txt file to point it to your installation of libtorch.

The easiest way to do that is to determine where your python package manager installed it.

Pip pip show torch Once you have that root directory for torch append onto it share/cmake/Torch/. This should be enough to work.

Running Instructions

---

From inside the build directory you can run the package with ./superPoint ../pretained/combinedSuperPoint.pt

NOTE

On the initial run you might have to edit line 220 of imageProcess.cc and set your local image path.