Stefano Esposito, Andreas Geiger
University of Tübingen, Autonomous Vision Group
⚠️ This is a work in progress research codebase designed with a focus on modularity; future updates *will try* not to disrupt existing functionalities.
Our goal is to provide a plug and play library to quickly develop and test new research ideas. We offer various data loaders for commonly used multi-view datasets in 3D reconstruction and view-synthesis, that work out of the box without further data processing.
Don't miss out on our documentation. It's still a work in progress, but together with our examples, it is the best way to get started.
Install the module from source using pip install .. More details here.
Checkout the state of currently supported datasets or work in progress here.
Our standardized camera uses the OpenCV camera coordinate system and it is described here.
Images taken from Andreas Geiger's Computer Vision lectures at the University of Tübingen.
# download data in ./data
bash scripts/download/nerf_synthetic.sh
# visualize dataset splits
python examples/dataset_splits_vis.py --scene-name lego data:nerf-synthetic
Functions located in any .deprecated folder may no longer work as expected. While they might be supported again in the future, this is not guaranteed.
This project is licensed under the Creative Commons Attribution 4.0 International License (CC BY 4.0). See the LICENSE file for details.
You are free to use, modify, and distribute this code as long as you provide proper attribution to the original author(s).
If you use this library for your research, please consider citing:
@misc{Esposito2024MVDatasets,
author = {Stefano Esposito and Andreas Geiger},
title = {MVDatasets: Standardized DataLoaders for 3D Computer Vision},
year = {2025},
url = {https://github.com/autonomousvision/mvdatasets},
note = {GitHub repository}
}