pith. sign in

arxiv: 2606.25278 · v1 · pith:IONZOS43new · submitted 2026-06-24 · 💻 cs.CV · cs.AI

Heterogeneous and Adept Snapshot Distillation for 3D Semantic Segmentation

Xiaopei Wu , Yuenan Hou , Junkai Xu , Wenxiao Wang , Binbin Lin , Yu Li , Ping Li , Haifeng Liu
show 2 more authors
Deng Cai Wanli Ouyang
This is my paper

Pith reviewed 2026-06-25 20:55 UTC · model grok-4.3

classification 💻 cs.CV cs.AI
keywords 3D semantic segmentationknowledge distillationmulti-modal fusionmodel ensemblingpoint cloudScanNetV2S3DIS
0
0 comments X

The pith

Distilling from multi-modal teachers and training snapshots raises accuracy of point-cloud 3D segmentation without extra inference cost.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper sets out to improve uni-modal point-cloud networks for 3D semantic segmentation by transferring knowledge from a stronger multi-modal teacher and from multiple model experts. It introduces two mechanisms: one that filters informative images for multi-modal fusion before distillation, and another that treats checkpoints saved during ordinary training as class-specific experts that supervise the student only on the classes where each checkpoint excels. The combined method reaches state-of-the-art numbers on ScanNetV2 and S3DIS while leaving the student’s inference speed and architecture unchanged. A sympathetic reader would care because the approach promises higher performance on existing 3D pipelines without the usual costs of extra sensors or separate ensemble training.

Core claim

Information-oriented Heterogeneous Distillation transfers complementary cues from a multi-modal teacher to the point-cloud student after selecting the most informative images from image sequences, while Adept Snapshot Distillation lets each freely available training snapshot supervise the student solely in the classes where that snapshot is strongest; the resulting HAS-KD framework produces state-of-the-art results on ScanNetV2 and S3DIS and integrates directly into contemporary point-cloud segmentation algorithms.

What carries the argument

HAS-KD, the combination of heterogeneous distillation from a multi-modal teacher and selective class-wise supervision from training snapshots treated as adept experts.

If this is right

  • Existing point-cloud segmentation networks can absorb multi-modal knowledge and ensemble-like benefits at training time only.
  • The student model attains higher mean intersection-over-union scores on standard indoor 3D benchmarks.
  • No auxiliary image inputs or additional model copies are required at inference.
  • The same training run that produces the teacher also supplies the expert snapshots, eliminating separate ensemble training.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The selective-supervision idea could be tested on other dense prediction tasks such as 3D instance segmentation or surface reconstruction.
  • If the adeptness criterion proves stable, it might reduce the need for explicit ensemble methods across a wider range of distillation settings.
  • The image-selection filter could be examined on outdoor datasets where image quality varies more sharply.

Load-bearing premise

Model snapshots created during ordinary training naturally form a collection of class-adept experts whose selective supervision improves the student without introducing harmful conflicts.

What would settle it

A controlled run in which the adept-snapshot supervision is applied to the student and either produces no accuracy gain or lowers accuracy on ScanNetV2 or S3DIS would falsify the central claim.

Figures

Figures reproduced from arXiv: 2606.25278 by Binbin Lin, Deng Cai, Haifeng Liu, Junkai Xu, Ping Li, Wanli Ouyang, Wenxiao Wang, Xiaopei Wu, Yuenan Hou, Yu Li.

Figure 1
Figure 1. Figure 1: Comparison with state-of-the-art algorithms in terms of performance and efficiency on ScanNetV2 val set. Our HAS-KD not [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Illustration of our Information-oriented Heterogeneous Distillation (IHD). It transfers the comprehensive knowledge of a multi [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Illustration of Information-Oriented Filtering (IOF) strategy. We select the view that holds the highest semantic information for [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Illustration of our Adept Snapshot Distillation (ASD). ASD transfers the knowledge of multiple expert teachers to the student. [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
read the original abstract

Multi-modal fusion and multi-model ensembling are prevalent in enhancing the performance of 3D semantic segmentation. Despite the impressive performance, these methods either rely on auxiliary input signals or suffer from costly computational expense. To efficaciously enhance the segmentation performance without introducing intolerable costs, we propose to transfer the rich knowledge from the multi-modal model (i.e., point clouds and images) and multiple model experts to the point-cloudbased network through knowledge distillation. Specifically, we present Information-oriented Heterogeneous Distillation (IHD) to help the uni-modal model absorb the complementary knowledge from the multi-modal teacher. We design the Information-Oriented Filtering (IOF) strategy to select informative images from the continuous image sequence for multi-modal fusion. This practice can boost the performance of the multi-modal teacher, thus benefiting the learning of the student. Besides, as opposed to vanilla model ensembling that requires the separate training of each expert, we propose Adept Snapshot Distillation (ASD). ASD treats the freely available model snapshots generated during the training phase as multiple experts, which significantly reduces the training cost for model ensembling. For each expert teacher, it only provides supervision to the student in the class where it is adept. The resulting Heterogeneous and Adept Snapshot Knowledge Distillation, dubbed HAS-KD, attains state-of-the-art results on ScanNetV2 and S3DIS datasets. HAS-KD can be seamlessly integrated into contemporary 3D segmentation algorithms and bring considerable gains without introducing extra inference burdens. The code will be made publicly available upon publication.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 3 minor

Summary. The manuscript introduces Heterogeneous and Adept Snapshot Knowledge Distillation (HAS-KD) for 3D semantic segmentation. It consists of Information-oriented Heterogeneous Distillation (IHD) that transfers complementary knowledge from a multi-modal (point-cloud + image) teacher to a point-cloud-only student, aided by Information-Oriented Filtering (IOF) to select informative images from sequences, together with Adept Snapshot Distillation (ASD) that treats ordinary training snapshots as class-adept experts and applies selective per-class supervision. The resulting method is reported to reach state-of-the-art performance on ScanNetV2 and S3DIS while adding no inference-time cost and being compatible with existing uni-modal pipelines.

Significance. If the reported gains hold, the work is significant for demonstrating a practical route to multi-modal and ensemble benefits without their usual deployment penalties. Explicit credit is due for supplying per-class accuracy curves of the snapshots, ablation tables that isolate the selective-supervision rule, and direct comparisons against non-selective ensembling baselines; these elements directly test the central assumption that snapshot experts can be used selectively without harmful conflicts. The stated intention to release code further supports reproducibility.

minor comments (3)
  1. [Abstract] Abstract: the claim of 'state-of-the-art results' would be more informative if the absolute mIoU numbers (or deltas) on ScanNetV2 and S3DIS were stated, even briefly.
  2. [§3.2] §3.2: the precise mathematical definition of the 'adeptness' selection rule (e.g., the threshold or ranking criterion applied to per-class accuracy) should be given as an equation rather than described only in prose.
  3. [Table 2] Table 2 (or equivalent ablation table): column headers and row labels should explicitly indicate whether the 'w/o selective' baseline uses the same snapshot teachers or a different ensembling strategy.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive assessment, the recognition of the work's significance, and the recommendation for minor revision. The report correctly notes the value of the per-class curves, ablations, and non-selective baselines in testing the core assumptions of ASD.

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper is an empirical algorithmic proposal for HAS-KD (combining IHD and ASD) that reports SOTA results on ScanNetV2 and S3DIS. No equations, fitted parameters, or derivation chain appear in the provided text. The method is described procedurally with ablation support and dataset comparisons; the central claim does not reduce to any self-definition, fitted-input prediction, or self-citation load-bearing step. This is the expected non-finding for a purely empirical contribution.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

No mathematical derivations, free parameters, or new postulated entities appear in the abstract; the contribution is algorithmic. Review is limited to the abstract alone.

pith-pipeline@v0.9.1-grok · 5836 in / 1091 out tokens · 23055 ms · 2026-06-25T20:55:33.371033+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

67 extracted references · 3 linked inside Pith

  1. [1]

    Zamir, Helen Jiang, Ioan- nis Brilakis, Martin Fischer, and Silvio Savarese

    Iro Armeni, Ozan Sener, Amir R. Zamir, Helen Jiang, Ioan- nis Brilakis, Martin Fischer, and Silvio Savarese. 3d seman- tic parsing of large-scale indoor spaces. InProceedings of the IEEE International Conference on Computer Vision and Pattern Recognition, 2016. 3, 5

    2016

  2. [2]

    Unstructured point cloud semantic labeling using deep segmentation networks.3dor@ eurographics, 3:17–24,

    Alexandre Boulch, Bertrand Le Saux, Nicolas Audebert, et al. Unstructured point cloud semantic labeling using deep segmentation networks.3dor@ eurographics, 3:17–24,

  3. [3]

    Largekernel3d: Scaling up kernels in 3d sparse cnns

    Yukang Chen, Jianhui Liu, Xiangyu Zhang, Xiaojuan Qi, and Jiaya Jia. Largekernel3d: Scaling up kernels in 3d sparse cnns. InProceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition, pages 13488–13498,

  4. [4]

    Segment and select: Vision-language segmentation in 3d scenarios.arXiv preprint arXiv:2606.10594, 2026

    Yulin Chen, Zhihang Zhong, and Yuenan Hou. Segment and select: Vision-language segmentation in 3d scenarios.arXiv preprint arXiv:2606.10594, 2026. 3

    Pith/arXiv arXiv 2026

  5. [5]

    Learning multi-modal trajectory poli- cies for data-efficient robotic manipulation.arXiv preprint arXiv:2606.01047, 2026

    Zijia Chen, Yuenan Hou, Xinhua Jiang, Yu Li, Weijie Li, and Li Liu. Learning multi-modal trajectory poli- cies for data-efficient robotic manipulation.arXiv preprint arXiv:2606.01047, 2026. 3

    Pith/arXiv arXiv 2026

  6. [6]

    4d spatio-temporal convnets: Minkowski convolutional neural networks

    Christopher Choy, JunYoung Gwak, and Silvio Savarese. 4d spatio-temporal convnets: Minkowski convolutional neural networks. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 3075–3084,

  7. [7]

    Pointcept: A codebase for point cloud perception research.https://github.com/ Pointcept/Pointcept, 2023

    Pointcept Contributors. Pointcept: A codebase for point cloud perception research.https://github.com/ Pointcept/Pointcept, 2023. 6

    2023

  8. [8]

    3dmv: Joint 3d-multi- view prediction for 3d semantic scene segmentation

    Angela Dai and Matthias Nießner. 3dmv: Joint 3d-multi- view prediction for 3d semantic scene segmentation. InPro- ceedings of the European Conference on Computer Vision (ECCV), pages 452–468, 2018. 3

    2018

  9. [9]

    Scannet: Richly-annotated 3d reconstructions of indoor scenes

    Angela Dai, Angel X Chang, Manolis Savva, Maciej Hal- ber, Thomas Funkhouser, and Matthias Nießner. Scannet: Richly-annotated 3d reconstructions of indoor scenes. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 5828–5839, 2017. 3, 5

    2017

  10. [10]

    Ensemble methods in machine learn- ing

    Thomas G Dietterich. Ensemble methods in machine learn- ing. InInternational workshop on multiple classifier systems, pages 1–15. Springer, 2000. 1, 3

    2000

  11. [11]

    3d semantic segmentation with submani- fold sparse convolutional networks

    Benjamin Graham, Martin Engelcke, and Laurens Van Der Maaten. 3d semantic segmentation with submani- fold sparse convolutional networks. InProceedings of the IEEE conference on computer vision and pattern recogni- tion, pages 9224–9232, 2018. 3

    2018

  12. [12]

    3d semantic segmentation with submani- fold sparse convolutional networks

    Benjamin Graham, Martin Engelcke, and Laurens Van Der Maaten. 3d semantic segmentation with submani- fold sparse convolutional networks. InProceedings of the IEEE conference on computer vision and pattern recogni- tion, pages 9224–9232, 2018. 1, 7

    2018

  13. [13]

    Neural network en- sembles.IEEE transactions on pattern analysis and machine intelligence, 12(10):993–1001, 1990

    Lars Kai Hansen and Peter Salamon. Neural network en- sembles.IEEE transactions on pattern analysis and machine intelligence, 12(10):993–1001, 1990. 1, 3

    1990

  14. [14]

    Distilling the Knowledge in a Neural Network.Statistics, 1050:9, 2015

    Geoffrey Hinton, Oriol Vinyals, and Jeff Dean. Distilling the Knowledge in a Neural Network.Statistics, 1050:9, 2015. 3

    2015

  15. [15]

    Learning lightweight lane detection CNNs by self atten- tion distillation

    Yuenan Hou, Zheng Ma, Chunxiao Liu, and Chen Change Loy. Learning lightweight lane detection CNNs by self atten- tion distillation. InIEEE International Conference on Com- puter Vision, pages 1013–1021, 2019. 3

    2019

  16. [16]

    Point-to-Voxel Knowledge Distillation for Li- DAR Semantic Segmentation

    Yuenan Hou, Xinge Zhu, Yuexin Ma, Chen Change Loy, and Yikang Li. Point-to-Voxel Knowledge Distillation for Li- DAR Semantic Segmentation. InIEEE Conference on Com- puter Vision and Pattern Recognition, pages 8479–8488,

  17. [17]

    Advances in 3d pre-training and downstream tasks: a survey.Vicinagearth, 1(1):6, 2024

    Yuenan Hou, Xiaoshui Huang, Shixiang Tang, Tong He, and Wanli Ouyang. Advances in 3d pre-training and downstream tasks: a survey.Vicinagearth, 1(1):6, 2024. 3

    2024

  18. [18]

    Bidirectional projection network for cross dimension scene understanding

    Wenbo Hu, Hengshuang Zhao, Li Jiang, Jiaya Jia, and Tien-Tsin Wong. Bidirectional projection network for cross dimension scene understanding. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 14373–14382, 2021. 1, 2, 3, 4, 7

    2021

  19. [19]

    Texturenet: Consistent local parametrizations for learning from high- resolution signals on meshes

    Jingwei Huang, Haotian Zhang, Li Yi, Thomas Funkhouser, Matthias Nießner, and Leonidas J Guibas. Texturenet: Consistent local parametrizations for learning from high- resolution signals on meshes. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4440–4449, 2019

    2019

  20. [20]

    Multi-view pointnet for 3d scene understanding

    Maximilian Jaritz, Jiayuan Gu, and Hao Su. Multi-view pointnet for 3d scene understanding. InProceedings of the IEEE/CVF International Conference on Computer Vision Workshops, pages 0–0, 2019. 3

    2019

  21. [21]

    Hierarchical point-edge interaction network for point cloud semantic segmentation

    Li Jiang, Hengshuang Zhao, Shu Liu, Xiaoyong Shen, Chi- Wing Fu, and Jiaya Jia. Hierarchical point-edge interaction network for point cloud semantic segmentation. InICCV,

  22. [22]

    Virtual multi-view fusion for 3d semantic segmentation

    Abhijit Kundu, Xiaoqi Yin, Alireza Fathi, David Ross, Brian Brewington, Thomas Funkhouser, and Caroline Pantofaru. Virtual multi-view fusion for 3d semantic segmentation. In ECCV, pages 518–535. Springer, 2020. 1, 7

    2020

  23. [23]

    Adaptive knowledge distillation based on entropy

    Kisoo Kwon, Hwidong Na, Hoshik Lee, and Nam Soo Kim. Adaptive knowledge distillation based on entropy. InICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 7409–7413. IEEE, 2020. 3

    2020

  24. [24]

    Stratified trans- former for 3d point cloud segmentation

    Xin Lai, Jianhui Liu, Li Jiang, Liwei Wang, Hengshuang Zhao, Shu Liu, Xiaojuan Qi, and Jiaya Jia. Stratified trans- former for 3d point cloud segmentation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 8500–8509, 2022. 6, 7

    2022

  25. [25]

    Deep projective 3d semantic segmentation

    Felix J ¨aremo Lawin, Martin Danelljan, Patrik Tosteberg, Goutam Bhat, Fahad Shahbaz Khan, and Michael Felsberg. Deep projective 3d semantic segmentation. InComputer Analysis of Images and Patterns: 17th International Confer- ence, CAIP 2017, Ystad, Sweden, August 22-24, 2017, Pro- ceedings, Part I 17, pages 95–107. Springer, 2017. 3

    2017

  26. [26]

    Seggcn: Effi- cient 3d point cloud segmentation with fuzzy spherical ker- nel

    Huan Lei, Naveed Akhtar, and Ajmal Mian. Seggcn: Effi- cient 3d point cloud segmentation with fuzzy spherical ker- nel. InCVPR, 2020. 7

    2020

  27. [27]

    Pointcnn: Convolution on x-transformed points.NeurIPS, 2018

    Yangyan Li, Rui Bu, Mingchao Sun, Wei Wu, Xinhan Di, and Baoquan Chen. Pointcnn: Convolution on x-transformed points.NeurIPS, 2018. 7 9

    2018

  28. [28]

    Moe3d: Mixture of ex- perts meets multi-modal 3d understanding.arXiv preprint arXiv:2511.22103, 2025

    Yu Li, Yuenan Hou, Yingmei Wei, Xinge Zhu, Yuexin Ma, Wenqi Shao, and Yanming Guo. Moe3d: Mixture of ex- perts meets multi-modal 3d understanding.arXiv preprint arXiv:2511.22103, 2025. 3

    arXiv 2025

  29. [29]

    Adaptive multi- teacher multi-level knowledge distillation.Neurocomputing, 415:106–113, 2020

    Yuang Liu, Wei Zhang, and Jun Wang. Adaptive multi- teacher multi-level knowledge distillation.Neurocomputing, 415:106–113, 2020. 3

    2020

  30. [30]

    Decoupled weight de- cay regularization

    Ilya Loshchilov and Frank Hutter. Decoupled weight de- cay regularization. InInternational Conference on Learning Representations (ICLR), 2018. 6

    2018

  31. [31]

    Rethinking network design and local geometry in point cloud: A simple residual mlp framework.arXiv preprint arXiv:2202.07123, 2022

    Xu Ma, Can Qin, Haoxuan You, Haoxi Ran, and Yun Fu. Rethinking network design and local geometry in point cloud: A simple residual mlp framework.arXiv preprint arXiv:2202.07123, 2022. 3

    arXiv 2022

  32. [32]

    V oxnet: A 3d con- volutional neural network for real-time object recognition

    Daniel Maturana and Sebastian Scherer. V oxnet: A 3d con- volutional neural network for real-time object recognition. In2015 IEEE/RSJ international conference on intelligent robots and systems (IROS), pages 922–928. IEEE, 2015. 3

    2015

  33. [33]

    Mix3d: Out-of-context data aug- mentation for 3d scenes

    Alexey Nekrasov, Jonas Schult, Or Litany, Bastian Leibe, and Francis Engelmann. Mix3d: Out-of-context data aug- mentation for 3d scenes. In2021 International Conference on 3D Vision (3DV), pages 116–125. IEEE, 2021. 6, 7

    2021

  34. [34]

    Popular ensemble meth- ods: An empirical study.Journal of artificial intelligence research, 11:169–198, 1999

    David Opitz and Richard Maclin. Popular ensemble meth- ods: An empirical study.Journal of artificial intelligence research, 11:169–198, 1999. 3

    1999

  35. [35]

    Fast point transformer

    Chunghyun Park, Yoonwoo Jeong, Minsu Cho, and Jae- sik Park. Fast point transformer. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 16949–16958, 2022. 7

    2022

  36. [36]

    Pointnet: Deep learning on point sets for 3d classification and segmentation

    Charles R Qi, Hao Su, Kaichun Mo, and Leonidas J Guibas. Pointnet: Deep learning on point sets for 3d classification and segmentation. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 652–660,

  37. [37]

    Pointnet++: Deep hierarchical feature learning on point sets in a metric space.Advances in neural information processing systems, 30, 2017

    Charles Ruizhongtai Qi, Li Yi, Hao Su, and Leonidas J Guibas. Pointnet++: Deep hierarchical feature learning on point sets in a metric space.Advances in neural information processing systems, 30, 2017. 3, 6, 7

    2017

  38. [38]

    Learn- ing multi-view aggregation in the wild for large-scale 3d se- mantic segmentation

    Damien Robert, Bruno Vallet, and Loic Landrieu. Learn- ing multi-view aggregation in the wild for large-scale 3d se- mantic segmentation. InProceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 5575–5584, 2022. 2, 3, 4, 7

    2022

  39. [39]

    FitNets: Hints for Thin Deep Nets

    Adriana Romero, Nicolas Ballas, Samira Ebrahimi Kahou, Antoine Chassang, Carlo Gatta, and Yoshua Bengio. FitNets: Hints for Thin Deep Nets. InInternational Conference on Learning Representations, 2015. 3

    2015

  40. [40]

    Smith and Nicholay Topin

    Leslie N. Smith and Nicholay Topin. Super-convergence: Very fast training of neural networks using large learning rates.arXiv preprint arXiv:1708.07120, 2017. 6

    Pith/arXiv arXiv 2017

  41. [41]

    Semantic scene com- pletion from a single depth image

    Shuran Song, Fisher Yu, Andy Zeng, Angel X Chang, Mano- lis Savva, and Thomas Funkhouser. Semantic scene com- pletion from a single depth image. InProceedings of the IEEE conference on computer vision and pattern recogni- tion, pages 1746–1754, 2017. 3

    2017

  42. [42]

    Tangent convolutions for dense prediction in 3d

    Maxim Tatarchenko, Jaesik Park, Vladlen Koltun, and Qian- Yi Zhou. Tangent convolutions for dense prediction in 3d. In CVPR, 2018. 7

    2018

  43. [43]

    Segcloud: Semantic segmen- tation of 3d point clouds

    Lyne Tchapmi, Christopher Choy, Iro Armeni, JunYoung Gwak, and Silvio Savarese. Segcloud: Semantic segmen- tation of 3d point clouds. In3DV, 2017. 6, 7

    2017

  44. [44]

    Kpconv: Flexible and deformable convolution for point clouds

    Hugues Thomas, Charles R Qi, Jean-Emmanuel Deschaud, Beatriz Marcotegui, Franc ¸ois Goulette, and Leonidas J Guibas. Kpconv: Flexible and deformable convolution for point clouds. InProceedings of the IEEE/CVF international conference on computer vision, pages 6411–6420, 2019. 7

    2019

  45. [45]

    Graph attention convolution for point cloud seman- tic segmentation

    Lei Wang, Yuchun Huang, Yaolin Hou, Shenman Zhang, and Jie Shan. Graph attention convolution for point cloud seman- tic segmentation. InCVPR, 2019. 7

    2019

  46. [46]

    Octformer: Octree-based transformers for 3d point clouds.arXiv preprint arXiv:2305.03045, 2023

    Peng-Shuai Wang. Octformer: Octree-based transformers for 3d point clouds.arXiv preprint arXiv:2305.03045, 2023. 1, 6, 7, 8

    arXiv 2023

  47. [47]

    O-cnn: Octree-based convolutional neu- ral networks for 3d shape analysis.ACM Transactions On Graphics (TOG), 36(4):1–11, 2017

    Peng-Shuai Wang, Yang Liu, Yu-Xiao Guo, Chun-Yu Sun, and Xin Tong. O-cnn: Octree-based convolutional neu- ral networks for 3d shape analysis.ACM Transactions On Graphics (TOG), 36(4):1–11, 2017. 7

    2017

  48. [48]

    Squeezeseg: Convolutional neural nets with recurrent crf for real-time road-object segmentation from 3d lidar point cloud

    Bichen Wu, Alvin Wan, Xiangyu Yue, and Kurt Keutzer. Squeezeseg: Convolutional neural nets with recurrent crf for real-time road-object segmentation from 3d lidar point cloud. In2018 IEEE international conference on robotics and au- tomation (ICRA), pages 1887–1893. IEEE, 2018. 3

    2018

  49. [49]

    Squeezesegv2: Improved model structure and unsupervised domain adaptation for road-object segmenta- tion from a lidar point cloud

    Bichen Wu, Xuanyu Zhou, Sicheng Zhao, Xiangyu Yue, and Kurt Keutzer. Squeezesegv2: Improved model structure and unsupervised domain adaptation for road-object segmenta- tion from a lidar point cloud. In2019 international confer- ence on robotics and automation (ICRA), pages 4376–4382. IEEE, 2019. 3

    2019

  50. [50]

    Multi-teacher knowledge distillation for compressed video action recognition on deep neural networks

    Meng-Chieh Wu, Ching-Te Chiu, and Kun-Hsuan Wu. Multi-teacher knowledge distillation for compressed video action recognition on deep neural networks. InICASSP 2019-2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 2202–2206. IEEE, 2019. 3

    2019

  51. [51]

    Point transformer v2: Grouped vector atten- tion and partition-based pooling.Advances in Neural Infor- mation Processing Systems, 35:33330–33342, 2022

    Xiaoyang Wu, Yixing Lao, Li Jiang, Xihui Liu, and Heng- shuang Zhao. Point transformer v2: Grouped vector atten- tion and partition-based pooling.Advances in Neural Infor- mation Processing Systems, 35:33330–33342, 2022. 1, 3, 7, 8

    2022

  52. [52]

    Point transformer v3: Simpler, faster, stronger.arXiv preprint arXiv:2312.10035, 2023

    Xiaoyang Wu, Li Jiang, Peng-Shuai Wang, Zhijian Liu, Xi- hui Liu, Yu Qiao, Wanli Ouyang, Tong He, and Hengshuang Zhao. Point transformer v3: Simpler, faster, stronger.arXiv preprint arXiv:2312.10035, 2023. 1, 3, 6, 7, 8

    arXiv 2023

  53. [53]

    Towards large- scale 3d representation learning with multi-dataset point prompt training.arXiv preprint arXiv:2308.09718, 2023

    Xiaoyang Wu, Zhuotao Tian, Xin Wen, Bohao Peng, Xihui Liu, Kaicheng Yu, and Hengshuang Zhao. Towards large- scale 3d representation learning with multi-dataset point prompt training.arXiv preprint arXiv:2308.09718, 2023. 7

    arXiv 2023

  54. [54]

    Taseg: Temporal aggregation network for lidar semantic segmentation

    Xiaopei Wu, Yuenan Hou, Xiaoshui Huang, Binbin Lin, Tong He, Xinge Zhu, Yuexin Ma, Boxi Wu, Haifeng Liu, Deng Cai, et al. Taseg: Temporal aggregation network for lidar semantic segmentation. InIEEE Conference on Com- puter Vision and Pattern Recognition, pages 15311–15320,

  55. [55]

    Paconv: Position adaptive convolution with dynamic ker- nel assembling on point clouds

    Mutian Xu, Runyu Ding, Hengshuang Zhao, and Xiaojuan Qi. Paconv: Position adaptive convolution with dynamic ker- nel assembling on point clouds. InCVPR, 2021. 7 10

    2021

  56. [56]

    Pointasnl: Robust point clouds processing using nonlocal neural networks with adaptive sampling

    Xu Yan, Chaoda Zheng, Zhen Li, Sheng Wang, and Shuguang Cui. Pointasnl: Robust point clouds processing using nonlocal neural networks with adaptive sampling. In Proceedings of the IEEE/CVF conference on computer vi- sion and pattern recognition, pages 5589–5598, 2020. 7

    2020

  57. [57]

    2dpass: 2d priors assisted semantic segmentation on lidar point clouds

    Xu Yan, Jiantao Gao, Chaoda Zheng, Chao Zheng, Ruimao Zhang, Shuguang Cui, and Zhen Li. 2dpass: 2d priors assisted semantic segmentation on lidar point clouds. In European Conference on Computer Vision, pages 677–695. Springer, 2022. 3

    2022

  58. [58]

    Towards efficient 3d object detection with knowl- edge distillation.Advances in Neural Information Process- ing Systems, 35:21300–21313, 2022

    Jihan Yang, Shaoshuai Shi, Runyu Ding, Zhe Wang, and Xi- aojuan Qi. Towards efficient 3d object detection with knowl- edge distillation.Advances in Neural Information Process- ing Systems, 35:21300–21313, 2022. 3

    2022

  59. [59]

    Sam3d: Segment anything in 3d scenes.arXiv preprint arXiv:2306.03908, 2023

    Yunhan Yang, Xiaoyang Wu, Tong He, Hengshuang Zhao, and Xihui Liu. Sam3d: Segment anything in 3d scenes.arXiv preprint arXiv:2306.03908, 2023. 2, 4

    arXiv 2023

  60. [60]

    Swin3d: A pretrained transformer backbone for 3d indoor scene understanding.arXiv preprint arXiv:2304.06906,

    Yu-Qi Yang, Yu-Xiao Guo, Jian-Yu Xiong, Yang Liu, Hao Pan, Peng-Shuai Wang, Xin Tong, and Baining Guo. Swin3d: A pretrained transformer backbone for 3d indoor scene understanding.arXiv preprint arXiv:2304.06906,

    arXiv

  61. [61]

    Learning from multiple teacher networks

    Shan You, Chang Xu, Chao Xu, and Dacheng Tao. Learning from multiple teacher networks. InProceedings of the 23rd ACM SIGKDD International Conference on Knowledge Dis- covery and Data Mining, pages 1285–1294, 2017. 3

    2017

  62. [62]

    Confidence- aware multi-teacher knowledge distillation

    Hailin Zhang, Defang Chen, and Can Wang. Confidence- aware multi-teacher knowledge distillation. InICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 4498–4502. IEEE, 2022. 3

    2022

  63. [63]

    Avatar knowledge distillation: Self- ensemble teacher paradigm with uncertainty.arXiv preprint arXiv:2305.02722, 2023

    Yuan Zhang, Weihua Chen, Yichen Lu, Tao Huang, Xiuyu Sun, and Jian Cao. Avatar knowledge distillation: Self- ensemble teacher paradigm with uncertainty.arXiv preprint arXiv:2305.02722, 2023. 3

    arXiv 2023

  64. [64]

    Pointweb: Enhancing local neighborhood features for point cloud processing

    Hengshuang Zhao, Li Jiang, Chi-Wing Fu, and Jiaya Jia. Pointweb: Enhancing local neighborhood features for point cloud processing. InProceedings of the IEEE/CVF con- ference on computer vision and pattern recognition, pages 5565–5573, 2019. 3, 7

    2019

  65. [65]

    Point transformer

    Hengshuang Zhao, Li Jiang, Jiaya Jia, Philip HS Torr, and Vladlen Koltun. Point transformer. InProceedings of the IEEE/CVF international conference on computer vision, pages 16259–16268, 2021. 6, 7

    2021

  66. [66]

    Boosting 3d object detection by simulating multimodality on point clouds

    Wu Zheng, Mingxuan Hong, Li Jiang, and Chi-Wing Fu. Boosting 3d object detection by simulating multimodality on point clouds. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 13638– 13647, 2022. 3

    2022

  67. [67]

    Unidistill: A universal cross-modality knowl- edge distillation framework for 3d object detection in bird’s- eye view

    Shengchao Zhou, Weizhou Liu, Chen Hu, Shuchang Zhou, and Chao Ma. Unidistill: A universal cross-modality knowl- edge distillation framework for 3d object detection in bird’s- eye view. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 5116– 5125, 2023. 3 11

    2023