Concerto: Joint 2D-3D Self-Supervised Learning Emerges Spatial Representations

Yujia Zhang, Xiaoyang Wu, Yixing Lao, Chengyao Wang, Zhuotao Tian, Naiyan Wang, Hengshuang Zhao · NeurIPS 2025 · 2025

Status: To read
Topics: backbones
Feeds: First paper · height sampling

What they did

Combines 3D self-distillation with a joint 2D–3D embedding for self-supervised point cloud learning; in linear probing the features beat standalone 2D and 3D self-supervised models.

My note

No note yet.

Code

Repository: Pointcept/Concerto
Framework: PyTorch
Licence: Apache-2.0
My status: Not tried

BibTeX

@inproceedings{zhang2025-concerto,
  title = {{Concerto: Joint 2D-3D Self-Supervised Learning Emerges Spatial Representations}},
  author = {Yujia Zhang and Xiaoyang Wu and Yixing Lao and Chengyao Wang and Zhuotao Tian and Naiyan Wang and Hengshuang Zhao},
  booktitle = {NeurIPS 2025},
  year = {2025},
  eprint = {2510.23607},
  archivePrefix = {arXiv}
}