This project provides a streamlined way to run COLMAP and 3D Gaussian Splatting (3DGS) pipelines in Google Colab. It enables users to perform 3D scene reconstruction from images directly in the cloud, making photogrammetry and neural graphics more accessible.
- COLMAP Integration: Automates the process of Structure-from-Motion (SfM) and Multi-View Stereo (MVS) using COLMAP, generating camera poses and sparse/dense point clouds from image sets.
- 3D Gaussian Splatting (3DGS): Quickly reconstructs and visualizes 3D scenes using recent 3DGS techniques, allowing high-quality and efficient scene rendering.
- Colab-friendly: No local GPU required. The notebooks are set up to use free or Pro Google Colab GPU resources.
- Easy-to-use Workflow: Simple instructions for uploading images, running the full pipeline, and visualizing the results.
- Open the Google Colab notebook provided in this repository.
- Upload your dataset of images or use provided sample datasets.
- Run the cells to:
- Install dependencies (COLMAP, 3DGS, visualization tools)
- Process images through COLMAP for camera and point cloud reconstruction
- Apply 3DGS for scene rendering and visualization
- Download or view the reconstructed 3D models/outputs.
- A Google account
- Access to Google Colab
- Images of your scene (JPG/PNG)
All dependencies are installed in the notebook—no local setup needed.
Visualized in PLAYCANVAS:
This repository is distributed under the MIT License.
Special thanks to the authors and contributors of COLMAP and the 3DGS community for open-sourcing their research and tools.