Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat
Cited by 158

The DOI matches its Crossref record; 13 of the 13 references it cites were found in a registry or an index.

CLIP2Scene: Towards Label-efficient 3D Scene Understanding by CLIP

Open access
Chen, Runnan, Liu, Youquan, Kong, Lingdong, Zhu, Xinge, Ma, Yuexin, Li, Yikang, Hou, Yuenan, Qiao, Yu, Wang, Wenping · 2023 · 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

Abstract

Contrastive Language-Image Pre-training (CLIP) achieves promising results in 2D zero-shot and few-shot learning. Despite the impressive performance in 2D, applying CLIP to help the learning in 3D scene understanding has yet to be explored. In this paper, we make the first attempt to investigate how CLIP knowledge benefits 3D scene understanding. We propose CLIP2Scene, a simple yet effective framework that transfers CLIP knowledge from 2D image-text pre-trained models to a 3D point cloud network. We show that the pre-trained 3D network yields impressive performance on various downstream tasks, i.e., annotation-free and fine-tuning with labelled data for semantic segmentation. Specifically, built upon CLIP, we design a Semantic-driven Cross-modal Contrastive Learning framework that pre-trains a 3D network via semantic and spatial-temporal consistency regularization. For the former, we first leverage CLIP's text semantics to select the positive and negative point samples and then employ the contrastive loss to train the 3D network. In terms of the latter, we force the consistency between the temporally coherent point cloud features and their corresponding image features. We conduct experiments on SemanticKITTI, nuScenes, and ScanNet. For the first time, our pre-trained network achieves annotation-free 3D semantic segmentation with 20.8% and 25.08% mIoU on nuScenes and ScanNet, respectively. When fine-tuned with 1% or 100% labelled data, our method significantly outperforms other self-supervised methods, with improvements of 8% and 1% mIoU, respectively. Furthermore, we demonstrate the generalizability for handling cross-domain datasets. Code is publicly available11https://github.com/runnanchen/CLIP2Scene..

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1109/cvpr52729.2023.00678

Cites

13 of 13 references verified0 citation errors · 0 not checkable

Statistics

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 13 references

Research works printed but not linked: 50

  1. chen, 0, A simple framework for contrastive learning of visual representations, In International Conference on Machine Learning, 1597
  2. yao, 2022, Detclip: Dictionary-enriched visual-concept paralleled pre-training for open-world detection, ArXiv Preprint
  3. chen, 2022, Towards 3d scene understanding by referring synthetic models, ArXiv Preprint
  4. 10.1109/ICCV48922.2021.01572 · DOI · Paper page
  5. 10.1109/CVPR46437.2021.01236 · DOI · Paper page
  6. zhang, 0, Point-clip: Point cloud understanding by clip, IEEE Conf Computer Vision and Pattern Recognition, 8552
  7. zablocki, 0, Context-aware zero-shot learning for object recognition, In International Conference on Machine Learning, 7292
  8. vaswani, 0, Attention is all you need, In Advances in Neural Information Processing Systems, 30
  9. tang, 0, Searching efficient 3d architectures with sparse point-voxel convolution, in European Conference on Computer Vision, 685
  10. chen, 2022, Zero-shot point cloud segmentation by transferring geometric primitives, ArXiv Preprint
  11. xie, 0, Pointcontrast: Unsupervised pretraining for 3d point cloud understanding, in European Conference on Computer Vision, 574
  12. chen, 2021, Referring self-supervised learning on 3d point cloud
  13. 10.1109/CVPR.2016.15 · DOI · Paper page
  14. 10.1007/s11263-022-01650-4 · DOI
  15. cheraghian, 2019, Mitigating the Hubness Problem for Zero-Shot Learning of 3D Objects, ArXiv Preprint
  16. 10.1109/CVPR.2019.00319 · DOI · Paper page
  17. 10.23919/MVA.2019.8758063 · DOI
  18. 10.1109/CVPR52688.2022.00966 · DOI
  19. rozenberszki, 0, Language-grounded indoor 3d semantic segmentation in the wild, in European Conference on Computer Vision, 125
  20. 10.1109/CVPRW.2018.00294 · DOI · Paper page
  21. 10.1109/3DV53792.2021.00107 · DOI
  22. qi, 0, Pointnet: Deep learning on point sets for 3d classification and segmentation, IEEE Conf Computer Vision and Pattern Recognition, 652
  23. 10.1109/TPAMI.2013.140 · DOI · Paper page
  24. liu, 2021, Learning from 2d: Contrastive pixel-to-point knowledge transfer for 3d pretraining, ArXiv Preprint
  25. 10.1609/aaai.v32i1.12244 · DOI
  26. 10.1109/CVPR52688.2022.01755 · DOI · Paper page
  27. 10.1109/CVPR.2016.575 · DOI · Paper page
  28. caron, 0, Unsupervised learning of visual features by contrasting cluster assignments, In Advances in Neural Information Processing Systems, 33, 9912
  29. 10.1109/CVPR42600.2020.00914 · DOI
  30. bucher, 0, Generating visual representations for zero-shot classification, In IEEE/CVF International Conference on Computer Vision Workshop, 2666
  31. 10.1109/CVPR52729.2023.02079 · DOI
  32. 10.1109/ICCV48922.2021.00583 · DOI · Paper page
  33. 10.1109/ICCV48922.2021.00647 · DOI · Paper page
  34. 10.1109/CVPR.2017.473 · DOI · Paper page
  35. kobayashi, 2022, Decomposing nerf for editing via feature field distillation, ArXiv Preprint
  36. 10.1109/CVPR52688.2022.00829 · DOI · Paper page
  37. hong, 2022, Unified 3d and 4d panoptic segmentation via dynamic shifting network, ArXiv Preprint
  38. 10.1109/CVPR.2015.7298911 · DOI · Paper page
  39. 10.1109/CVPR.2013.111 · DOI · Paper page
  40. 10.1109/ICRA48891.2023.10160410 · DOI
  41. kong, 2023, Rethinking range view representation for lidar segmentation, ArXiv Preprint
  42. 10.1109/LRA.2022.3148457 · DOI · Paper page
  43. 10.1609/aaai.v29i1.9800 · DOI
  44. 10.1109/CVPR.2017.261 · DOI · Paper page
  45. 10.1109/ICCV.2017.139 · DOI
  46. 10.1109/TPAMI.2020.3005434 · DOI · Paper page
  47. han, 0, Co-teaching: Robust training of deep neural networks with extremely noisy labels, Advances in neural information processing systems, 31
  48. 10.1109/ICCV48922.2021.01009 · DOI · Paper page
  49. 10.1109/CVPR46437.2021.00981 · DOI · Paper page
  50. zhou, 0, Extract free dense labels from clip, in European Conference on Computer Vision, 696

Cited by

See all 158

Metadata

Authors
Chen, R., Liu, Y., Kong, L., Zhu, X., Ma, Y., Li, Y., Hou, Y., Qiao, Y., & Wang, W.
Venue
2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
Publisher
IEEE
Published
2023-06
Pages
7020-7030
ISSN
1063-6919 (ISSN-L)
OpenAlex
W4386075580
Cited by
158
References
13
FWCI
29.87
Type
conference paper
Language
en
Open access
green

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: CLIP2Scene: Towards Label-efficient 3D Scene Understanding by CLIP. https://app.cochat.ai/papers/10.1109/cvpr52729.2023.00678
The paper
Chen, R., Liu, Y., Kong, L., Zhu, X., Ma, Y., Li, Y., Hou, Y., Qiao, Y., & Wang, W. (2023). CLIP2Scene: Towards Label-efficient 3D Scene Understanding by CLIP. 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). https://doi.org/10.1109/cvpr52729.2023.00678

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.