tsdf-fusion

Why we included this project

Researchers working with RGB-D data often need to fuse a set of registered depth frames into a single volumetric model, and this CUDA/C++ project is a compact implementation of the classic projective TSDF approach. It integrates the depth maps into a truncated signed distance function voxel volume and exports surface point clouds, with a MATLAB script for generating meshes you can open in Meshlab. The same fusion code sits inside the 3DMatch toolbox, so anyone building reconstruction or scene-understanding pipelines will recognize the exact formulation and can adapt it rather than reimplementing from scratch. It is an older codebase, tested on Ubuntu 14.04 and 16.04 with OpenCV 2.4, so plan on adapting the build for a current system; the author also maintains a Python port with CPU and GPU support if you want something easier to run.