NDR-code
The project introduces Neural-DynamicReconstruction (NDR), a method for reconstructing dynamic scenes with high fidelity using a monocular RGB-D camera. NDR is a template-free approach that accurately captures geometry, motion, and appearance. The project includes comprehensive setup guides, data conventions, and pre-processing steps. Built on NeuS technology, it integrates dependencies like PyMCubes and opencv_python for optimal results. It provides tools for training, evaluating models, and rendering geometries. Presented at NeurIPS 2022, this research advances dynamic scene visualization, serving both academic and industrial purposes.
