BlenderNeuralangelo

Blender addon to inspect and preprocess COLMAP data for Neuralangelo (CVPR 2023). The addon allows you to inspect the COLMAP poses, check the sparse reconstruction overlaid with RGB images, and export the bounding regions for Neuralangelo.

Getting Started

First clone the code to a local directory.

cd /path/to/dir
git clone git@github.com:mli0603/BlenderNeuralangelo.git

0: Running COLMAP

Please follow the instructions to obtain COLMAP results. Alternatively, an example is provided in the folder toy_example.

If COLMAP instructions are followed correctly, a folder structure of the following should be expected:

DATA_PATH
├─ database.db      (COLMAP database)
├─ images           (undistorted input images)
├─ images_raw       (raw input images)
├─ sparse           (COLMAP data from SfM)
│  ├─ cameras.bin   (camera parameters)
│  ├─ images.bin    (images and camera poses)
│  ├─ points3D.bin  (sparse point clouds)
│  ├─ 0             (a directory containing individual SfM models. There could also be 1, 2... etc.)
│  ...
├─ stereo (COLMAP data for MVS, not used here)
...

1.1: Run Blender from command line

You can directly run Blender without worrying installation by:

./blender --python PATH_TO_BlenderNeuralangelo/start_blender_with_addon.py

Replace PATH_TO_BlenderNeuralangelo with your path to BlenderNeuralangelo.

1.2: Install as Addon

Installing as a Blender Addon avoids re-installation after quitting blender. To install as an Addon:

1.3: Run as code

Alternatively, the code can be directly run as follows:

After quitting Blender, this step must repeat.

2: Locating the control panel

After installation, the BlenderNeuralangelo panel can be located on the right side or press N on the keyboard.

3: Load COLMAP data

COLMAP can be loaded by providing the path to the COLMAP work directory. The images, camera parameters and sparse reconstruction results are loaded.

4: Inspect COLMAP data

COLMAP data can be inspected qualitatively by looking at the RGB, sparse reconstruction and pose results.

5: Defining region of interest

A graphics interface is provided to define the region of interest for Neuralangelo as a form of bounding box. Points outside the region of interest can be cropped.

6: Defining bounding sphere

Neuralangelo and prior works ¹²³ on neural surface reconstruction often assumes a spherical region of interest. This addon generates a bounding sphere and exports the scene parameters (intrinsics, poses, bounding sphere) following the Instant NGP format.