Welcome to the official repository for Vocabulary-Free 3D Instance Segmentation with Vision and Language Assistant! This repository hosts the source code, documentation, and resources for the PoVo project.
- The paper has been accepted by 3DV 2025 🔥.
- We have released the main model code and are currently fine-tuning it. Stay tuned for updates!
PoVo is a 3D Instance Segmentation method that can identify all object instances in a point cloud while associating them with semantic labels, without relying on any predefined vocabulary.
Visit the PoVo website to explore more details about the project, methodology, and results.
We welcome and encourage contributions to the PoVo project! If you'd like to contribute:
- Fork this repository.
- Create a new branch for your changes.
- Submit a pull request with a detailed description of your modifications.
Before contributing, please review our contribution guidelines.
We extend our gratitude to all contributors and supporters of the PoVo project. Your valuable insights and contributions drive innovation and progress in the field of 3D and language-based AI systems.
For questions, issues, or collaboration opportunities:
- Submit a ticket on the issues page.
- Visit the PoVo project website.
- Alternatively, reach out via email: gmei@fbk.eu.
This project is licensed under the Creative Commons Attribution-ShareAlike 4.0 International License.
For more information, visit the Creative Commons License page.