Related references
Note: Only part of the references are listed.
Article
Computer Science, Information Systems
Yunqian He et al.
Summary: Existing point cloud classification researches are usually conducted on complete and semantically clear datasets. However, in real point cloud scenes, occlusion and truncation can affect classification performance by destroying object completeness. To address this issue, we propose an incomplete point cloud classification network (IPC-Net) that utilizes data augmentation and similarity measurement. IPC-Net learns feature representation of incomplete point clouds and semantic differences compared to complete ones for improved classification. Experimental results validate the ability of IPC-Net to classify incomplete point clouds and enhance the robustness of point cloud classification under varying completeness levels.
JOURNAL OF VISUAL COMMUNICATION AND IMAGE REPRESENTATION
(2023)
Article
Computer Science, Artificial Intelligence
Maciej Zamorski et al.
Summary: Contemporary deep neural networks show excellent performance in visual reasoning, especially in the context of 3D point cloud data. However, processing this type of data remains challenging. This study proposes a novel neural network architecture capable of continual learning on 3D point cloud data and reduces catastrophic forgetting by utilizing the point cloud structure properties. By using rehearsal and reconstruction as regularization methods, our approach significantly decreases catastrophic forgetting compared to existing solutions on multiple popular point cloud datasets in both known and unknown task settings.
COMPUTER VISION AND IMAGE UNDERSTANDING
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Xiuwei Xu et al.
Summary: In this paper, the authors propose BSC-Net, a binary sparse convolutional network, for efficient point cloud analysis. By searching for the optimal subset of convolution operation, the quantization errors in sparse convolution can be significantly alleviated, resulting in improved network performance.
2023 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, CVPR
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Sneha Paul et al.
Summary: In this study, we introduce a novel self-supervised learning method called CrossMoCo, which utilizes 2D rendered images of point clouds in a multi-modal setup for representation learning. CrossMoCo outperforms existing methods on multi-modal self-supervised learning, achieving improved performance on point cloud tasks.
2023 20TH CONFERENCE ON ROBOTS AND VISION, CRV
(2023)
Article
Computer Science, Artificial Intelligence
Huafeng Wang et al.
Summary: This study addresses the challenge of pose variations in object classification based on point cloud by developing a novel end-to-end pose robust graph convolutional network. Experimental results show that the new model outperforms existing approaches when conducting experiments on random rotations of 3D point clouds in ModelNet40 and ShapeNetCore datasets.
PATTERN RECOGNITION
(2022)
Article
Computer Science, Artificial Intelligence
Chen Zhao et al.
Summary: In this paper, a method for achieving rotation invariance in point cloud analysis is proposed by combining local geometry with global topology. Experimental results demonstrate that the method achieves state-of-the-art performance on various datasets.
PATTERN RECOGNITION
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Sneha Paul et al.
Summary: This research work is based on the PointNet architecture and aims to improve the accuracy of the PointNet model. The ModelNet10 dataset is used, and variations of encoder models, improved training protocol, and transfer learning from larger datasets are proposed. The experiments show a 6.10% improvement over the baseline model.
STRUCTURAL, SYNTACTIC, AND STATISTICAL PATTERN RECOGNITION, S+SSPR 2022
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Jaesung Choe et al.
Summary: MLP-Mixer is a new model that shows noticeable performance in image recognition tasks using channel-mixing and token-mixing. However, it faces limitations when applied to sparse and unordered point clouds. To overcome this, we propose PointMixer, a universal point set operator that allows feature mixing within and between point sets.
COMPUTER VISION - ECCV 2022, PT XXVII
(2022)
Article
Computer Science, Artificial Intelligence
Yiming Cui et al.
Summary: This paper proposes a geometric attentional dynamic graph convolutional neural network for point cloud analysis, which can extract both intrinsic and extrinsic properties of point clouds for rich representation learning of point features. By modeling the relations in geometric space as geometric attention and incorporating them in EdgeConv, the network is able to capture the intrinsic feature likelihood of point clouds.
Article
Computer Science, Software Engineering
Meng-Hao Guo et al.
Summary: This paper introduces a novel framework named Point Cloud Transformer (PCT) for point cloud learning, based on Transformer and enhanced by farthest point sampling and nearest neighbor search for better capturing local context. Extensive experiments demonstrate that the PCT achieves state-of-the-art performance on shape classification, part segmentation, semantic segmentation, and normal estimation tasks.
COMPUTATIONAL VISUAL MEDIA
(2021)
Article
Computer Science, Artificial Intelligence
Anh Viet Phan et al.