Filters








105,597 Hits in 8.5 sec

Learning point embedding for 3D data processing [article]

Zhenpeng Chen, Yuan li
2021 arXiv   pre-print
By analysis of PointNet, a pioneer in introducing deep learning into point sets, we reveal that current point-based methods are essentially spatial relationship processing networks.  ...  Our architecture, named PE-Net, learns the representation of point clouds in high-dimensional space, and encodes the unordered input points to feature vectors, which standard 2D CNNs can be applied to.  ...  We believe that our idea of learning point embedding can be potentially generalized to 3D related tasks. We leave as our future work. Fig. 1 . 1 Fig. 1.  ... 
arXiv:2107.08565v2 fatcat:xzbuezc6jrabfddfncixtqziuy

CpT: Convolutional Point Transformer for 3D Point Cloud Processing [article]

Chaitanya Kaul, Joshua Mitton, Hang Dai, Roderick Murray-Smith
2021 arXiv   pre-print
We present CpT: Convolutional point Transformer - a novel deep learning architecture for dealing with the unstructured nature of 3D point cloud data.  ...  It achieves this feat due to its effectiveness in creating a novel and robust attention-based point set embedding through a convolutional projection layer crafted for processing dynamically local point  ...  Deep learning progress on processing 3D data was initially slow primarily due to the fact that early deep learning required a structured input data representation as a prerequisite.  ... 
arXiv:2111.10866v1 fatcat:qp7ccsxfpzhd3fxbkxn4t5dhbm

SAWNet: A Spatially Aware Deep Neural Network for 3D Point Cloud Processing [article]

Chaitanya Kaul, Nick Pears, Suresh Manandhar
2019 arXiv   pre-print
But their application to processing data lying on non-Euclidean domains is still a very active area of research.  ...  These techniques use either global or local information from the point clouds to extract a latent representation for the points, which is then used for the task at hand (classification/segmentation).  ...  VoxNet [21] sample 3D points into a 32 × 32 × 32 3D occupancy grid which is processed by a 3D CNN.  ... 
arXiv:1905.07650v1 fatcat:ckyraevd3vftve7ylickbbqdse

TreeGCN-ED: Encoding Point Cloud using a Tree-Structured Graph Network [article]

Prajwal Singh, Kaustubh Sadekar, Shanmuganathan Raman
2022 arXiv   pre-print
Point cloud is one of the widely used techniques for representing and storing 3D geometric data. In the past several methods have been proposed for processing point clouds.  ...  We further demonstrate the applicability of the proposed framework in applications like: 3D point cloud completion and Single image-based 3D reconstruction.  ...  Also, we would like to thank Ashish Tiwari and Dhananjay Singh for their constructive and valuable feedback.  ... 
arXiv:2110.03170v3 fatcat:tukv2b6i7fbtzdo37yzz4yz4fu

MeshMAE: Masked Autoencoders for 3D Mesh Data Analysis [article]

Yaqian Liang, Shanshan Zhao, Baosheng Yu, Jing Zhang, Fazhi He
2022 arXiv   pre-print
In this paper, we explore this learning paradigm for 3D mesh data analysis based on Transformers.  ...  Then, through reconstructing the information of masked patches, the network is capable of learning discriminative representations for mesh data.  ...  For 3D data analysis, PointBERT [64] generalizes the concept of BERT into 3D point clouds by devising a masked point reconstruction task to pre-train the Transformers.  ... 
arXiv:2207.10228v1 fatcat:5dvuiqhyhzatrcfbfa3odq27pe

Deep Learning-Based 3D Instance and Semantic Segmentation: A Review

Siddiqui Muhammad Yasir, Hyunsik Ahn
2022 Journal on Artificial Intelligence  
The process of segmenting point cloud data into several homogeneous areas with points in the same region having the same attributes is known as 3D segmentation.  ...  However, due to the specific problems of processing point clouds with deep neural networks, deep learning on point clouds is still in its initial stages.  ...  Large-scale point clouds require data pre-processing to deal with problems like these.  ... 
doi:10.32604/jai.2022.031235 fatcat:nsxfjoy4mnamfkrsu2mgsoums4

Spatial Semantic Embedding Network: Fast 3D Instance Segmentation with Deep Metric Learning [article]

Dongsu Zhang, Junha Chun, Sang Kyun Cha, Young Min Kim
2020 arXiv   pre-print
We propose spatial semantic embedding network (SSEN), a simple, yet efficient algorithm for 3D instance segmentation using deep metric learning.  ...  For high-level intelligent tasks from a large scale scene, 3D instance segmentation recognizes individual instances of objects.  ...  D. 3D Point Cloud Instance Segmentation Compared to the recognition or segmentation for images, the instance segmentation in 3D data has acquired its attention lately with the recent advance of 3D deep  ... 
arXiv:2007.03169v1 fatcat:cms4hzgwjnf6hkqaeclr4eln6y

3D Instance Segmentation via Multi-Task Metric Learning

Jean Lahoud, Bernard Ghanem, Martin R. Oswald, Marc Pollefeys
2019 2019 IEEE/CVF International Conference on Computer Vision (ICCV)  
We propose a novel method for instance label segmentation of dense 3D voxel grids 1 .  ...  We solve the 3D instance-labeling problem with a multi-task learning strategy.  ...  Government is authorized to reproduce and distribute reprints for Governmental purposes notwithstanding any copyright annotation thereon.  ... 
doi:10.1109/iccv.2019.00935 dblp:conf/iccv/LahoudGOP19 fatcat:scpoy35slrdp3ihglqa2c6a6ya

SFGAN: Unsupervised Generative Adversarial Learning of 3D Scene Flow from the 3D Scene Self

Guangming Wang, Chaokang Jiang, Zehang Shen, Yanzi Miao, Hesheng Wang
2021 Advanced Intelligent Systems  
The flow embedding layer learns a flow embedding d k between PC 1 and PC 2 for each point in PC 1 .  ...  For example, 'PC 1 ⇒ PC Ã 2 ' means finding the softly corresponding points in PC Ã 2 for each point in PC 1 and learning the flow embedding for each point in PC 1 .  ...  Data Availability Statement Research data are not shared.  ... 
doi:10.1002/aisy.202100197 fatcat:5poimouzazeqhf2xaf4ze5wtxi

Visual Analytics for Deep Embeddings of Large Scale Molecular Dynamics Simulations [article]

Junghoon Chae, Debsindhu Bhowmik, Heng Ma, Arvind Ramanathan, Chad Steed
2019 bioRxiv   pre-print
Processing this immense amount of data in a meaningful way is becoming increasingly difficult.  ...  However, most of the existing visualizations for embeddings have limitations in evaluating the embedding models and understanding the complex simulation data.  ...  A. 3D Embedding View After dimensionality reduction, we are given a set of data points in the 3D latent space. Our 3D embedding view displays the 3D data points as circle dots as shown in Figure 2 .  ... 
doi:10.1101/830844 fatcat:fhmqbdal5jdn7btzmcb5pqeihm

3D Vision with Transformers: A Survey [article]

Jean Lahoud, Jiale Cao, Fahad Shahbaz Khan, Hisham Cholakkal, Rao Muhammad Anwer, Salman Khan, Ming-Hsuan Yang
2022 arXiv   pre-print
We discuss transformer design in 3D vision, which allows it to process data with various 3D representations.  ...  Although a number of surveys have focused on transformers in vision in general, 3D vision requires special attention due to the difference in data representation and processing when compared to 2D vision  ...  In this section, we first review different representations of 3D data, as well as numerous processing techniques that enable learning from such data.  ... 
arXiv:2208.04309v1 fatcat:h7xk3hydevhwhatx5lil3ftnri

3D Instance Segmentation via Multi-Task Metric Learning [article]

Jean Lahoud, Bernard Ghanem, Marc Pollefeys, Martin R. Oswald
2019 arXiv   pre-print
We propose a novel method for instance label segmentation of dense 3D voxel grids.  ...  We solve the 3D instance-labeling problem with a multi-task learning strategy.  ...  Our approach is robust and scalable, therefore suitable for processing the large amounts of data in a 3D scene. • Our experiments demonstrate state-of-the-art performance for 3D instance segmentation.  ... 
arXiv:1906.08650v2 fatcat:h43y73r7yjho3oihbhnofdx3k4

Unsupervised 3D Reconstruction from a Single Image via Adversarial Learning [article]

Lingjing Wang, Yi Fang
2017 arXiv   pre-print
Recent advancements in deep learning opened new opportunities for learning a high-quality 3D model from a single 2D image given sufficient training on large-scale data sets.  ...  Then, we jointly train a 3D deconvolutional network to transform the latent vector space to the 3D object space together with the embedding process.  ...  On one hand, for most supervised methods, such as T-L embedding [17] , 3D-R2N2 [19] , 3DGAN [18] and 3D Point Clouds Method [20] , robust performance requires a large number of labeled data for training  ... 
arXiv:1711.09312v1 fatcat:epf33ijd75axnmcmcxmew3u5yq

A Convolutional Architecture for 3D Model Embedding [article]

Arniel Labrada, Benjamin Bustos, Ivan Sipiran
2021 arXiv   pre-print
Our ex-periments show the benefit of computing the embeddings of a 3D modeldata set and use them for effective 3D Model Retrieval.  ...  Since this vector is a learned repre-sentation which tries to capture the relevant information of a 3D model,we show that the embedding representation conveys semantic informationthat helps to deal with  ...  A survey on this topic can be found in Deep Learning for 3D Point Clouds: A Survey [12] .  ... 
arXiv:2103.03764v1 fatcat:pnnbs577kzhlpmu2mhulohivfe

Efficient 3D Deep LiDAR Odometry [article]

Guangming Wang, Xinrui Wu, Shuyang Jiang, Zhe Liu, Hesheng Wang
2022 arXiv   pre-print
An efficient 3D point cloud learning architecture, named EfficientLO-Net, for LiDAR odometry is first proposed in this paper.  ...  In this architecture, the projection-aware representation of the 3D point cloud is proposed to organize the raw 3D point cloud into an ordered data form to achieve efficiency.  ...  This 3D filtering operation is similar to the point cloud post-processing method of RangeNet++ [47] , but we integrate this process into the feature learning process for 3D point clouds.  ... 
arXiv:2111.02135v2 fatcat:kf5q4f7n6ngtjkol5lcounet2m
« Previous Showing results 1 — 15 out of 105,597 results