This repository contains the official implementation of the ECCV paper "McGrids: Monte Carlo-Driven Adaptive Grids for Iso-Surface Extraction". McGrids introduces a novel approach to enhance the efficiency of iso-surface extraction by constructing adaptive grids, as opposed to the traditional uniform grids used in prior work.
McGrids leverages a Monte Carlo process to solve the problem of constructing adaptive grids as a probability sampling problem. The result is a significant reduction in the number of implicit field queries, which leads to substantial memory savings while producing high-quality meshes with detailed geometry. We validate McGrids through extensive experiments, including both analytical signed distance functions (SDFs) from surface meshes and learned implicit fields from real multiview images.
Before proceeding with the installation, ensure that you have installed the following dependency:
- Geogram: Please follow the instructions provided in the Geogram repository to install it. After installation, update the path to Geogram in
./differentiable_mcmt/CMakeLists.txt.
To set up the environment and install McGrids, follow these steps:
conda create -n mcgrids python=3.10.0
conda activate mcgrids
pip install .To extract a mesh from an SDF using McGrids, you can run the following example:
cd examples
python extract_mesh_from_sdf.py --input_path ../assets/armadillo.obj --output_path ../assets/armadillo_mcgrids.obj