VisRecon

VisRecon is a research-oriented toolkit for capturing RGB-D imagery, preparing multi-view datasets, running COLMAP reconstruction workflows, and inspecting the resulting point clouds or meshes. It combines small command-line utilities with sample outputs from experiments using an Intel RealSense camera.

Overview

The repository supports a compact multi-view reconstruction workflow: capture or download image sets, prepare them for reconstruction, run a configurable COLMAP pipeline, and visualize the generated model. Samples and rendered demonstrations are included to make the expected inputs and outputs easier to understand.

Features

• Capture RGB-D sequences from supported Intel RealSense cameras
• Prepare color and depth images for reconstruction experiments
• Download selected folders from a Hugging Face dataset repository
• Run a scripted COLMAP reconstruction pipeline
• Visualize sparse or dense point clouds and meshes with Open3D
• Review included sample images and reconstruction previews

Structure

.
├── demo/          # Rendered reconstruction examples
├── samples/       # Sample multi-view images
├── utils/         # Capture, download, preparation, and visualization tools
├── requirements.txt
└── LICENSE

Requirements

• Python 3.10 or later
• COLMAP available from the command line
• A supported Intel RealSense device for capture workflows

Install the Python dependencies:

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

On Windows, activate the environment with .venv\Scripts\activate.

Dataset Download

Download a folder from a Hugging Face dataset repository by supplying its repository ID:

python utils/dataset_downloader.py \
  --repo_id username/dataset \
  --subset path/to/subset \
  --local_dir ./data

Usage

Capture images

python utils/image_fetcher.py --help

Prepare images

python utils/image_preparer.py --help

Run reconstruction

Review the paths and options near the beginning of utils/run_colmap_pipeline.sh, then run:

bash utils/run_colmap_pipeline.sh

Visualize a model

python utils/visualizer.py --help

Samples

The samples directory contains a representative input sequence. The demo directory contains rendered mesh and point-cloud results for plant reconstructions.

Project Status

Experimental. VisRecon is retained as a research toolkit and reference workflow. Its utilities require local hardware, datasets, and COLMAP configuration and should not be treated as a production reconstruction system.

Contributing

Focused bug fixes, documentation improvements, and reusable reconstruction utilities are welcome. Open an issue before proposing a substantial workflow change.

License

The source code in this repository is available under the MIT License. Dataset and third-party tool licenses remain separate and should be reviewed at their respective sources.

Contact

For questions or collaboration, use the contact details below or consult the website for the latest information.

• Website: yixuanhuang.com
• Email: yixnhuang@gmail.com