An optimized, lightweight CUDA-based 3D reconstruction implementation derived from the original Kinfu algorithm.
KinFu remake is an optimized and reworked implementation of the original Kinfu algorithm for real-time 3D reconstruction from depth sensors. It reconstructs 3D models using depth camera input with improved performance and cleaner code architecture. The project solves the need for a more efficient and maintainable version of the classic Kinfu algorithm while removing dependencies on specific libraries.
Computer vision researchers and developers working with real-time 3D reconstruction, particularly those using CUDA-enabled GPUs and depth sensing hardware like Kinect.
Developers choose this implementation for its 1.6x performance improvement over the original, cleaner codebase with runtime parameter configuration, and independence from specific library dependencies like PCL and OpenCV GPU module.
Optimized and reworked version of Kinfu
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Delivers up to 1.6x faster execution on Fermi architecture, enhancing real-time 3D scanning efficiency for depth sensor inputs.
Codebase is drastically smaller and more readable, making it easier to modify and extend compared to the original PCL implementation.
All algorithm parameters like volume size can be adjusted at runtime without hardcoded limits, allowing flexible experimentation.
Independent from OpenCV GPU module and PCL library, simplifying integration and reducing potential conflicts in standalone systems.
Requires specific versions of CUDA 5.0+, OpenCV 2.4.9 with Viz module (and VTK), and OpenNI v1.5.4, making installation and setup error-prone.
Limited to NVIDIA GPUs with Fermi or newer architecture, excluding other GPU vendors and potentially outdated for modern hardware.
Relies on OpenNI v1.5.4, which may not support newer depth sensors like Kinect v2 without additional modifications or drivers.