KinectFusion is an approach for reconstructing rigid geometric shapes using a consumer-grade RGB-D camera. This repository is a development repository for creating an OpenCL implementation of the approach.
Although most of the dependencies can be automatically downloaded using the setup_external.sh script, the project requires OpenCL to be installed correctly.
The Project currently depends on the OopenCV library wnd its contributers's modules that can be automatically downloaded and installed using the setup_external.sh script in the project folder. The following libraries are installed from the ubuntu software repository or downloaded, built from source and installed into the libs directory:
- OpenCV with extra modules from OpenCV Contrib
The project is able to read the assoociated rgb and depth files from all sequences in the RGB-D SLAM Dataset and Benchmark. Furthermore it provides download scripts downloadFreiburg.py for the fr1/xyz dataset that also performs frame association. The dataset is downloaded to the data folder.
The project is simply built with the command cmake . in the project directory followed by make. Note that building in a separate build folder requres copying the OpenCL kernels manually to <build-folder>/src/cl.
Executing the project with ./KinectFusion_OpenCL will run the reconstruction on the downloaded dataset.