TUN3D: Towards Real-World Scene Understanding from Unposed Images
📰 News
- :fire: January 2026 — TUN3D is accepted to ICRA 2026!
- :fire: 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.
<div align="center"> <video src="https://github.com/user-attachments/assets/8644a6d7-3a4e-4b1b-b58e-023276ea12ee"> </video> <p><i>TUN3D works with GT point clouds, posed images (with known camera poses), or fully unposed image sets (without poses or depths).</i></p> </div>TUN3D: Towards Real-World Scene Understanding from Unposed Images<br> Anton Konushin Nikita Drozdov, Bulat Gabdullin, Alexey Zakharov, Anna Vorontsova, Danila Rukhovich, Maksim Kolodiazhnyi <br> https://arxiv.org/abs/2509.21388
Installation
The repository is divided into two modules:
<p float="left"> <img src="recognition/imgs/predictions_scannet.png" width="900" height="396" /> </p> <p float="left"> <img src="recognition/imgs/predictions_s3dis.png" width="900" height="396" /> </p>