Research on Consistency-Driven Point Cloud Semantic Segmentation Method in Label Noise Scenarios

Bo Xia; Lili Wang

doi:10.63313/JCSFT.9058

Authors

Bo Xia College of Physics and Electronic Engineering, Northwest Normal University, Lanzhou 730070, PR China Author
Lili Wang College of Physics and Electronic Engineering, Northwest Normal University, Lanzhou 730070, PR China Author

DOI:

https://doi.org/10.63313/JCSFT.9058

Keywords:

Point cloud semantic segmentation, Noisy labels, Consistency learning, Robust training t,

Abstract

In real-world scenarios, manual annotation of point cloud data often involves significant noise, which severely degrades the training performance of semantic segmentation models. To address this issue, this paper proposes a consistency-driven point cloud semantic segmentation framework that jointly leverages augmentation consistency and historical consistency strategies to dynamically filter training samples and refine pseudo-labels, thereby improving robustness under noisy supervision. Specifically, we design a multi-layer encoder–decoder network based on local feature aggregation and attention pooling to effectively capture the geometric structure and semantic contextual information of point clouds. During training, multiple augmented predictions are generated for the same input, and the reliability of pseudo-labels is assessed based on prediction confidence and stability. Meanwhile, the historical prediction trajectories of samples are incorporated to further filter and correct pseudo-labels, enhancing the stability of the supervision signal. Under simulated noisy-label settings on the S3DIS dataset, the proposed method achieves overall accuracies (OA) of 85.16% and 82.66% at noise rates of 60% and 80%, respectively, outperforming mainstream baseline methods. Experimental results demonstrate the effectiveness of consistency constraints in noisy-label scenarios and provide a new perspective for robust learning in point cloud semantic segmentation.

References

[1] Han, B., Yao, Q., Yu, X., Niu, G., Xu, M., Hu, W., ... & Sugiyama, M. (2018). Co-teaching: Robust training of deep neural networks with extremely noisy labels. Advances in neural information processing systems, 31.

[2] Li, J., Socher, R., & Hoi, S. C. (2020). Dividemix: Learning with noisy labels as semi-supervised learning. arXiv preprint arXiv:2002.07394.

[3] Liu, S., Niles-Weed, J., Razavian, N., & Fernandez-Granda, C. (2020). Early-learning regularization prevents memorization of noisy labels. Advances in neural information processing systems, 33, 20331-20342.

[4] Tarvainen, A., & Valpola, H. (2017). Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results. Advances in neural information processing systems, 30.

[5] Sohn, K., Berthelot, D., Carlini, N., Zhang, Z., Zhang, H., Raffel, C. A., ... & Li, C. L. (2020). Fixmatch: Simplifying semi-supervised learning with consistency and confidence. Advances in neural information processing systems, 33, 596-608.

[6] Chen, T., Kornblith, S., Norouzi, M., & Hinton, G. (2020, April). Simclr: A simple framework for contrastive learning of visual representations. In International Conference on Learning Representations (Vol. 2, No. 4). New York, NY, USA: PMLR.

[7] Qi, C. R., Su, H., Mo, K., & Guibas, L. J. (2017). Pointnet: Deep learning on point sets for 3d classification and segmentation. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 652-660).

[8] Qi, C. R., Yi, L., Su, H., & Guibas, L. J. (2017). Pointnet++: Deep hierarchical feature learning on point sets in a metric space. Advances in neural information processing systems, 30.

[9] Wang, Y., Sun, Y., Liu, Z., Sarma, S. E., Bronstein, M. M., & Solomon, J. M. (2019). Dynamic graph cnn for learning on point clouds. ACM Transactions on Graphics (tog), 38(5), 1-12.

[10] Thomas, H., Qi, C. R., Deschaud, J. E., Marcotegui, B., Goulette, F., & Guibas, L. J. (2019). Kpconv: Flexible and deformable convolution for point clouds. In Proceedings of the IEEE/CVF international conference on computer vision (pp. 6411-6420).

[11] Hu, Q., Yang, B., Xie, L., Rosa, S., Guo, Y., Wang, Z., ... & Markham, A. (2020). Randla-net: Efficient semantic segmentation of large-scale point clouds. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition (pp. 11108-11117).

[12] Zhao, H., Jiang, L., Jia, J., Torr, P. H., & Koltun, V. (2021). Point transformer. In Proceedings of the IEEE/CVF international conference on computer vision (pp. 16259-16268).

[13] Lai, X., Liu, J., Jiang, L., Wang, L., Zhao, H., Liu, S., ... & Jia, J. (2022). Stratified transformer for 3d point cloud segmentation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition (pp. 8500-8509).

[14] Ye, S., Chen, D., Han, S., & Liao, J. (2021). Learning with noisy labels for robust point cloud segmentation. In Proceedings of the IEEE/CVF international conference on computer vision (pp. 6443-6452).