Filters








185 Hits in 4.6 sec

One Framework to Register Them All: PointNet Encoding for Point Cloud Alignment [article]

Vinit Sarode, Xueqian Li, Hunter Goforth, Yasuhiro Aoki, Animesh Dhagat, Rangaprasad Arun Srivatsan, Simon Lucey, Howie Choset
2019 arXiv   pre-print
This paper presents a novel framework that uses PointNet encoding to align point clouds and perform registration for applications such as 3D reconstruction, tracking and pose estimation.  ...  We develop a framework that compares PointNet features of template and source point clouds to find the transformation that aligns them accurately.  ...  point clouds using PointNet as an encoding function.  ... 
arXiv:1912.05766v1 fatcat:kfmlo4dilva6bbiatc3i2yh22a

PCRNet: Point Cloud Registration Network using PointNet Encoding [article]

Vinit Sarode, Xueqian Li, Hunter Goforth, Yasuhiro Aoki, Rangaprasad Arun Srivatsan, Simon Lucey, Howie Choset
2019 arXiv   pre-print
We develop a framework that compares PointNet features of template and source point clouds to find the transformation that aligns them accurately.  ...  This paper presents a novel framework that uses the PointNet representation to align point clouds and perform registration for applications such as tracking, 3D reconstruction and pose estimation.  ...  Our approach uses PointNet in a Siamese architecture to encode the shape information of a template and a source point cloud as feature vectors, and estimates the pose that aligns these two features using  ... 
arXiv:1908.07906v2 fatcat:5s7smvw7k5e2fazb7ihgse5jxe

PointNetLK: Robust & Efficient Point Cloud Registration using PointNet [article]

Yasuhiro Aoki, Hunter Goforth, Rangaprasad Arun Srivatsan, Simon Lucey
2019 arXiv   pre-print
to point cloud registration.  ...  As a consequence, classical vision algorithms for image alignment can be applied on the problem - namely the Lucas & Kanade (LK) algorithm.  ...  of points, (2) sensitive to initialization, and (3) nontrivial to integrate them to deep learning framework due to issues of differentiability.  ... 
arXiv:1903.05711v2 fatcat:3honraepobf2no7i5f5gmjv2za

Deep Global Features for Point Cloud Alignment

Ahmed El Khazari, Yue Que, Thai Leang Sung, Hyo Jong Lee
2020 Sensors  
The proposed model consisted of three main parts: an encoding network, an auxiliary module that weighed the contribution of each input point cloud, and feature alignment to achieve the final transform.  ...  However, despite the fact that point cloud registration or alignment is addressed in learning-based methods, such as PointNetLK, they do not offer good generalizability for point clouds.  ...  To address this representation of the point cloud, the PointNet network was utilized as an encoding function. φ denotes the PointNet function as φ : R 3×N −→ R K , so that the input point cloud is P ∈  ... 
doi:10.3390/s20144032 pmid:32698504 pmcid:PMC7411762 fatcat:qejzwml62bhznmpmxc4hvzynsm

When Deep Learning Meets Data Alignment: A Review on Deep Registration Networks (DRNs)

Victor Villena-Martinez, Sergiu Oprea, Marcelo Saval-Calvo, Jorge Azorin-Lopez, Andres Fuster-Guillo, Robert B. Fisher
2020 Applied Sciences  
In this review, we classify these methods according to a proposed framework based on the traditional registration pipeline.  ...  This pipeline consists of four steps: target selection, feature extraction, feature matching, and transform computation for the alignment.  ...  Deep Closest Point [70] registers two point clouds by first embedding them into high-dimensional space using DGCNN [101] to extract features.  ... 
doi:10.3390/app10217524 fatcat:zevtfrfsuzhyliknzg5thaprbq

Efficient Learning on Point Clouds with Basis Point Sets [article]

Sergey Prokudin, Christoph Lassner, Javier Romero
2019 arXiv   pre-print
In a second experiment, we show how the proposed representation can be used for registering high-resolution meshes to noisy 3D scans.  ...  Using the proposed representation as the input to a simple fully connected network allows us to match the performance of PointNet on a shape classification task while using three orders of magnitude less  ...  To find correspondences between two point clouds, we process each of them with our network, obtaining as a result two registered mesh templates.  ... 
arXiv:1908.09186v1 fatcat:5cmnv3okkneohof2ysbptrge54

Automatic Bone Surface Restoration for Markerless Computer-Assisted Orthopaedic Surgery

Xue Hu, Ferdinando Rodriguez y Baena
2022 Chinese Journal of Mechanical Engineering  
The bone surface modification is inevitable due to intra-operative intervention. The mismatched correspondences will degrade the reliability of registered target pose.  ...  According to the evaluation on both synthetic data and real-time captures, the registration quality can be effectively improved by surface reconstruction.  ...  PointNet, based on the same encoder-decoder design, is regarded as the most popular backbone for point cloud-based learning [18] .  ... 
doi:10.1186/s10033-022-00684-6 fatcat:eybziekw3vbshntk47ih2cpb4y

SCRnet: A Spatial Consistency Guided Network Using Contrastive Learning for Point Cloud Registration

Huixiang Shao, Zhijiang Zhang, Xiaoyu Feng, Dan Zeng
2022 Symmetry  
Point cloud registration is used to find a rigid transformation from the source point cloud to the target point cloud.  ...  Therefore, to address this issue, we propose a spatial consistency guided network using contrastive learning for point cloud registration (SCRnet), in which its overall stage is symmetrical.  ...  Due to carrying out feature learning on the global point cloud, PointNet cannot obtain local features, which makes it difficult to analyze complex scenes.  ... 
doi:10.3390/sym14010140 fatcat:wclxljmrjrc3nfcwjtvrcy566u

Multispectral LiDAR Point Cloud Classification Using SE-PointNet++

Zhuangwei Jing, Haiyan Guan, Peiran Zhao, Dilong Li, Yongtao Yu, Yufu Zang, Hanyun Wang, Jonathan Li
2021 Remote Sensing  
The SE block is embedded into PointNet++ to strengthen important channels to increase feature saliency for better point cloud classification.  ...  In this paper, a point-wise multispectral LiDAR point cloud classification architecture termed as SE-PointNet++ is proposed via integrating a Squeeze-and-Excitation (SE) block with an improved PointNet  ...  Acknowledgments: The authors would like to thank all the colleagues for the fruitful discussions on this work.  ... 
doi:10.3390/rs13132516 fatcat:ffm2ht54bjh65pjf6qudzuoshe

Drought Stress Classification using 3D Plant Models [article]

Siddharth Srivastava, Swati Bhugra, Brejesh Lall, Santanu Chaudhury
2017 arXiv   pre-print
In this context, an accurate 3D model of plant canopy provides a reliable representation for drought stress characterization in contrast to using 2D images.  ...  To overcome the high degree of self-similarities and self-occlusions in plant canopy, prior knowledge of leaf shape based on features from deep siamese network are used to construct an accurate 3D model  ...  PointNet generates a global feature on the input point cloud.  ... 
arXiv:1709.09496v2 fatcat:seaxwxo4wjd27onzsayd4myseq

Active 3D Classification of Multiple Objects in Cluttered Scenes

Yiming Wang, Marco Carletti, Francesco Setti, Marco Cristani, Alessio Del Bue
2019 2019 IEEE/CVF International Conference on Computer Vision Workshop (ICCVW)  
Among many, accurate object classification is an essential supporting element for assistive robotics.  ...  By reducing the impact of occlusions, we show with both synthetic and real-world data that in a few moves the approach can surpass a state-of-the-art method, Point-Net with single view object classification  ...  Together with the registered points, ICP also provides the pose transform ∆p j m (t) that is applied to align the object model to the point cloud segment.  ... 
doi:10.1109/iccvw.2019.00318 dblp:conf/iccvw/WangCSCB19 fatcat:5mmqe7sjhbaf3iuykwwyzzbvse

Drought Stress Classification Using 3D Plant Models

Siddharth Srivastava, Swati Bhugra, Brejesh Lall, Santanu Chaudhury
2017 2017 IEEE International Conference on Computer Vision Workshops (ICCVW)  
In this context, an accurate 3D model of plant canopy provides a reliable representation for drought stress characterization in contrast to using 2D images.  ...  To overcome the high degree of self-similarities and selfocclusions in plant canopy, prior knowledge of leaf shape based on features from deep siamese network are used to construct an accurate 3D model  ...  PointNet generates a global feature on the input point cloud.  ... 
doi:10.1109/iccvw.2017.240 dblp:conf/iccvw/SrivastavaBLC17 fatcat:m4esh2je2ff2xj4why2nnmykbm

When Deep Learning Meets Data Alignment: A Review on Deep Registration Networks (DRNs) [article]

Victor Villena-Martinez, Sergiu Oprea, Marcelo Saval-Calvo, Jorge Azorin-Lopez, Andres Fuster-Guillo, Robert B. Fisher
2020 arXiv   pre-print
Commonly, a registration process can be divided into four main steps: target selection, feature extraction, feature matching, and transform computation for the alignment.  ...  We classify the different papers proposing a framework extracted from the traditional registration pipeline to analyse the new learning-based proposal strengths.  ...  This work has also been supported by two Valencian Government grants for PhD studies (ACIF/2017/223 and ACIF/2018/197).  ... 
arXiv:2003.03167v1 fatcat:5diw5hkahjg5ndayvt4ne7zl2y

Learning an Effective Equivariant 3D Descriptor Without Supervision [article]

Riccardo Spezialetti and Samuele Salti and Luigi Di Stefano
2019 arXiv   pre-print
path that goes in the opposite direction of end-to-end learning from raw data so successfully deployed for 2D images.  ...  The effectiveness of the proposed approach is experimentally validated by outperforming hand-crafted and learned descriptors on a standard benchmark.  ...  Learning from Raw 3D Data PointNet [21] and Point-Net++ [22] are pioneering works presenting a general framework to learn features directly from raw point clouds data.  ... 
arXiv:1909.06887v1 fatcat:afohotttqjcwjkagpprgvq5jue

PointNetLK Revisited [article]

Xueqian Li, Jhony Kaesemodel Pontes, Simon Lucey
2021 arXiv   pre-print
Hybrid learning methods, that use learning for predicting point correspondences and then a deterministic step for alignment, have offered some respite, but are still limited in their generalization abilities  ...  We address the generalization ability of recent learning-based point cloud registration methods.  ...  Acknowledgments: The authors would like to thank Ioannis Gkioulekas, Ming-Fang Chang, Haosen Xing, and the reviewers for their invaluable suggestions that significantly improved this work.  ... 
arXiv:2008.09527v2 fatcat:omgv7uorxval3eiku5onzugr2q
« Previous Showing results 1 — 15 out of 185 results