TUN3D: Towards Real-World Scene Understanding from Unposed Images

📰 News

• 🔥 September 2025 — Initial release of TUN3D!

This repository contains an implementation of TUN3D, a method for real-world indoor scene understanding from multi-view images.

teaser.mp4

TUN3D works with GT point clouds, posed images (with known camera poses), or fully unposed image sets (without poses or depths).

TUN3D: Towards Real-World Scene Understanding from Unposed Images
Anton Konushin Nikita Drozdov, Bulat Gabdullin, Alexey Zakharov, Anna Vorontsova, Danila Rukhovich, Maksim Kolodiazhnyi
https://arxiv.org/abs/2509.21388

Installation

The repository is divided into two modules:

1. Reconstruction
2. Recognition

Each module requires a separate installation of dependencies. Please follow the installation guide provided in each module’s directory.

Data preprocessing

• Preprocessing instructions and scripts are located in the corresponding folders: Scannet, S3DIS, Structured3d.
• All preprocessed datasets are also available on Hugging Face. The installation guide provides detailed steps on how to download them correctly.

Running

After completing the data preprocessing stage, navigate to the recognition folder and follow the instructions provided there.

Predictions example

ScanNet

S3DIS

Citation

If you find this work useful for your research, please cite our paper:

@misc{konushin2025tun3drealworldsceneunderstanding,
      title={TUN3D: Towards Real-World Scene Understanding from Unposed Images}, 
      author={Anton Konushin and Nikita Drozdov and Bulat Gabdullin and Alexey Zakharov and Anna Vorontsova and Danila Rukhovich and Maksim Kolodiazhnyi},
      year={2025},
      eprint={2509.21388},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2509.21388}, 
}