3,142 Hits in 5.5 sec

Guided Point Contrastive Learning for Semi-supervised Point Cloud Semantic Segmentation [article]

Li Jiang, Shaoshuai Shi, Zhuotao Tian, Xin Lai, Shu Liu, Chi-Wing Fu, Jiaya Jia
2021 arXiv   pre-print
To address the high cost and challenges of 3D point-level labeling, we present a method for semi-supervised point cloud semantic segmentation to adopt unlabeled point clouds in training to boost the model  ...  Inspired by the recent contrastive loss in self-supervised tasks, we propose the guided point contrastive loss to enhance the feature representation and model generalization ability in semi-supervised  ...  Pseudo Guidance on Contrastive Learning Now, we focus on the setting of semi-supervised learning (SSL) for 3D point cloud semantic segmentation, in which we could leverage some labeled data to train the  ... 
arXiv:2110.08188v1 fatcat:i25xkcemj5hopn6pgicf4infue

Teachers in concordance for pseudo-labeling of 3D sequential data [article]

Awet Haileslassie Gebrehiwot, Patrik Vacek, David Hurych, Karel Zimmermann, Patrick Perez, Tomáš Svoboda
2022 arXiv   pre-print
The output of multiple teachers is combined via a novel pseudo-label confidence-guided criterion. Our experimental evaluation focuses on the 3D point cloud domain in urban driving scenarios.  ...  We show the performance of our method applied to multiple model architectures with tasks of 3D semantic segmentation and 3D object detection on two benchmark datasets.  ...  A semisupervised learning to 3D semantic segmentation with guided point contrastive loss has been proposed in [12] .  ... 
arXiv:2207.06079v1 fatcat:xeeu6sthxnco5ishc33g5htmdi

Data Efficient 3D Learner via Knowledge Transferred from 2D Model [article]

Ping-Chung Yu, Cheng Sun, Min Sun
2022 arXiv   pre-print
Collecting and labeling the registered 3D point cloud is costly. As a result, 3D resources for training are typically limited in quantity compared to the 2D images counterpart.  ...  Specifically, we utilize a strong and well-trained semantic segmentation model for 2D images to augment RGB-D images with pseudo-label. The augmented dataset can then be used to pre-train 3D models.  ...  Existing works have explored self-supervised pre-training and semi-supervised learning using unlabeled scene point clouds to improve performance.  ... 
arXiv:2203.08479v2 fatcat:4xhrrwld7ngs3kz4b6ry6ba364

PriFit: Learning to Fit Primitives Improves Few Shot Point Cloud Segmentation [article]

Gopal Sharma and Bidya Dash and Aruni RoyChowdhury and Matheus Gadelha and Marios Loizou and Liangliang Cao and Rui Wang and Erik Learned-Miller and Subhransu Maji and Evangelos Kalogerakis
2022 arXiv   pre-print
We present PriFit, a semi-supervised approach for label-efficient learning of 3D point cloud segmentation networks.  ...  The learned point representations can then be re-used in existing network architectures for 3D point cloud segmentation, and improves their performance in the few-shot setting.  ...  Conclusion We propose a simple semi-supervised learning approach, PRIFIT, for learning point embeddings for few shot semantic segmentation.  ... 
arXiv:2112.13942v2 fatcat:cpp6bjosjrgu7kk7avsmitukda

SESS: Self-Ensembling Semi-Supervised 3D Object Detection

Na Zhao, Tat-Seng Chua, Gim Hee Lee
2020 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)  
The performance of existing point cloud-based 3D object detection methods heavily relies on large-scale highquality 3D annotations.  ...  Semi-supervised learning is a good alternative to mitigate the data annotation issue, but has remained largely unexplored in 3D object detection.  ...  Semi-supervised learning is a promising alternative to strongly supervised learning for point cloud-based 3D object detection.  ... 
doi:10.1109/cvpr42600.2020.01109 dblp:conf/cvpr/ZhaoCL20 fatcat:ljq5yrkb5ngt3l4ogxyijw4bce

SESS: Self-Ensembling Semi-Supervised 3D Object Detection [article]

Na Zhao, Tat-Seng Chua, Gim Hee Lee
2021 arXiv   pre-print
The performance of existing point cloud-based 3D object detection methods heavily relies on large-scale high-quality 3D annotations.  ...  Semi-supervised learning is a good alternative to mitigate the data annotation issue, but has remained largely unexplored in 3D object detection.  ...  Semi-supervised learning is a promising alternative to strongly supervised learning for point cloud-based 3D object detection.  ... 
arXiv:1912.11803v3 fatcat:7tmdcso3bbfcpobzjho2hlhega

Scribble-Supervised LiDAR Semantic Segmentation [article]

Ozan Unal and Dengxin Dai and Luc Van Gool
2022 arXiv   pre-print
In this paper, we propose using scribbles to annotate LiDAR point clouds and release ScribbleKITTI, the first scribble-annotated dataset for LiDAR semantic segmentation.  ...  Our pipeline comprises of three stand-alone contributions that can be combined with any LiDAR semantic segmentation model to achieve up to 95.7% of the fully-supervised performance while using only 8%  ...  Acknowledgements: Special thanks to Zeynep Demirkol and Tim Brödermann for their efforts during annotation.  ... 
arXiv:2203.08537v2 fatcat:jwqvpolcfvg7pmaxrvkm34xcwa

Reconstruction-Aware Prior Distillation for Semi-supervised Point Cloud Completion [article]

Zhaoxin Fan, Yulin He, Zhicheng Wang, Kejian Wu, Hongyan Liu, Jun He
2022 arXiv   pre-print
In contrast, this paper proposes a novel Reconstruction-Aware Prior Distillation semi-supervised point cloud completion method named RaPD, which takes advantage of a two-stage training scheme to reduce  ...  In training stage 1, the so-called deep semantic prior is learned from both unpaired complete and unpaired incomplete point clouds using a reconstruction-aware pretraining process.  ...  [6] propose to divide a point cloud into superpoints and build superpoint graphs for semi-supervised point cloud segmentation.  ... 
arXiv:2204.09186v2 fatcat:g6cuoay72fcwhnwjqupnj7rwu4

Unsupervised Representation Learning for 3D Point Cloud Data [article]

Jincen Jiang, Xuequan Lu, Wanli Ouyang, Meili Wang
2021 arXiv   pre-print
Though a number of point cloud learning methods have been proposed to handle unordered points, most of them are supervised and require labels for training.  ...  By contrast, unsupervised learning of point cloud data has received much less attention to date. In this paper, we propose a simple yet effective approach for unsupervised point cloud learning.  ...  Index Terms-unsupervised contrastive learning, point cloud, 3D object classification, semantic segmentation. I.  ... 
arXiv:2110.06632v1 fatcat:y7aqgedxgja4fg2d4vdcnhweoe

Object Segmentation for Autonomous Driving Using iseAuto Data

Junyi Gu, Mauro Bellone, Raivo Sell, Artjom Lind
2022 Electronics  
Following this line of research, this paper approaches the problem of object segmentation using LiDAR–camera fusion and semi-supervised learning implemented in a fully convolutional neural network.  ...  In this work, it is shown that with LiDAR–camera fusion, with only a few annotated scenarios and semi-supervised learning, it is possible to achieve robust performance on real-world data in a multi-class  ...  learning for semantic segmentation.  ... 
doi:10.3390/electronics11071119 fatcat:dv4jvexqbfdqjnwiq3qyovdtmy

One Million Scenes for Autonomous Driving: ONCE Dataset [article]

Jiageng Mao, Minzhe Niu, Chenhan Jiang, Hanxue Liang, Jingheng Chen, Xiaodan Liang, Yamin Li, Chaoqiang Ye, Wei Zhang, Zhenguo Li, Jie Yu, Hang Xu (+1 others)
2021 arXiv   pre-print
To facilitate future research on exploiting unlabeled data for 3D detection, we additionally provide a benchmark in which we reproduce and evaluate a variety of self-supervised and semi-supervised methods  ...  However, the research community generally suffered from data inadequacy of those essential real-world scene data, which hampers the future exploration of fully/semi/self-supervised methods for 3D perception  ...  For self-supervised learning, PointContrast [48] and DepthContrast [56] apply contrastive learning on point clouds.  ... 
arXiv:2106.11037v3 fatcat:fwgrb57yarhujmetzpewtdzzei

SLiDE: Self-supervised LiDAR De-snowing through Reconstruction Difficulty [article]

Gwangtak Bae, Byungjun Kim, Seongyong Ahn, Jihong Min, Inwook Shim
2022 arXiv   pre-print
To address this problem, we propose a novel self-supervised learning framework for snow points removal in LiDAR point clouds.  ...  Semantic segmentation with snow labels would be a straightforward solution for removing them, but it requires laborious point-wise annotation.  ...  It takes advantage of deep learning-based semantic segmentation methods to detect point-wise LiDAR noise points.  ... 
arXiv:2208.04043v1 fatcat:tetapfpuozdy5eqcophgsifoai

Label-Efficient Point Cloud Semantic Segmentation: An Active Learning Approach [article]

Xian Shi, Xun Xu, Ke Chen, Lile Cai, Chuan Sheng Foo, Kui Jia
2021 arXiv   pre-print
Deep learning models are the state-of-the-art methods for semantic point cloud segmentation, the success of which relies on the availability of large-scale annotated datasets.  ...  cloud segmentation.  ...  To reduce these annotation costs, recent work has focused on label-efficient learning strategies for point cloud segmentation, including semi-supervised learning and unsupervised learning [31, 26] .  ... 
arXiv:2101.06931v2 fatcat:awmakdxm5bawhaimd55wosgkxe

PointWise: An Unsupervised Point-wise Feature Learning Network [article]

Matan Shoef, Sharon Fogel, Daniel Cohen-Or
2019 arXiv   pre-print
By clustering the learned embedding space, we perform unsupervised part-segmentation on point clouds. By calculating euclidean distance in the latent space we derive semantic point-analogies.  ...  We present a novel approach to learning a point-wise, meaningful embedding for point-clouds in an unsupervised manner, through the use of neural-networks.  ...  We have shown an implementation of the embedding learned for part-segmentation of point-clouds.  ... 
arXiv:1901.04544v2 fatcat:wheuhu7vwbedvndc5tdi6ddec4

One Thing One Click: A Self-Training Approach for Weakly Supervised 3D Semantic Segmentation [article]

Zhengzhe Liu, Xiaojuan Qi, Chi-Wing Fu
2021 arXiv   pre-print
Point cloud semantic segmentation often requires largescale annotated training data, but clearly, point-wise labels are too tedious to prepare.  ...  Experimental results on both ScanNet-v2 and S3DIS show that our self-training approach, with extremely-sparse annotations, outperforms all existing weakly supervised methods for 3D semantic segmentation  ...  Related Work Semantic Segmentation for Point Cloud Approaches for 3D semantic segmentation can be roughly divided into pointbased methods and voxel-based methods.  ... 
arXiv:2104.02246v4 fatcat:ucniyqpo6zgi5k4ned7my5nafq
« Previous Showing results 1 — 15 out of 3,142 results