PPFNet: Global Context Aware Local Features for Robust 3D Point Matching

By Haowen Deng, Tolga Birdal, Slobodan Ilic. 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2018

639 citations recorded by the source corpus.

We present PPFNet - Point Pair Feature NETwork for deeply learning a globally informed 3D local feature descriptor to find correspondences in unorganized point clouds. PPFNet learns local descriptors on pure geometry and is highly aware of the global context, an important cue in deep learning. Our 3D representation is computed as a collection of point-pair-features combined with the points and normals within a local vicinity. Our permutation invariant network design is inspired by PointNet and sets PPFNet to be ordering-free. As opposed to voxelization, our method is able to consume raw point clouds to exploit the full sparsity. PPFNet uses a novel N-tuple loss and architecture injecting the global information naturally into the local descriptor. It shows that context awareness also boosts the local feature representation. Qualitative and quantitative evaluations of our network suggest increased recall, improved robustness and invariance as well as a vital step in the 3D descriptor extraction performance.

Subfields: Aerospace Engineering, Computational Mechanics, Geology

View Paper: https://www.semanticscholar.org/paper/04258f92294d5dea89adb3e8607ae22eadb5332d

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Aerospace Engineering

In Computational Mechanics

In Geology

Citations in the corpus

PREDATOR: Registration of 3D Point Clouds with Low Overlap (2020, background): “PPFNet [10], and its self-supervised version PPFFoldNet [9], combine point pair features with a PointNet [28] architecture to extract descriptors that are aware of the global context.”
Geometric Transformer for Fast and Robust Point Cloud Registration (2022, background): “The recent advances have been dominated by learningbased, correspondence-based methods [4,7,9,14,15,36].”
CoFiNet: Reliable Coarse-to-fine Correspondences for Robust Point Cloud Registration (2021, methodology): “Differently, the last category of methods [6, 7, 8, 9, 10, 11, 12] tackles point cloud registration in a two-stage manner.”
SuperGlue: Learning Feature Matching With Graph Neural Networks (2019, background): “Some works treat all elements equally, through global pooling [62, 37, 13] or instance normalization [54, 30, 29], while others focus on a local neighborhood in coordinate or feature space [38, 60].”
L3-Net: Towards Learning Based LiDAR Localization for Autonomous Driving (2019, methodology): “Some existing methods [41, 7, 10] could also be applied to solve the re-localization problem with point clouds.”
SpinNet: Learning a General Surface Descriptor for 3D Point Cloud Registration (2020, methodology): “Second, although a number of recent approaches [14, 13, 22, 37] introduce rotation-invariant descriptors, they simply integrate the handcrafted features such as point-pairs [48, 16] and point density [50, 62, 51], or external local reference frames (LRFs) [37, 33, 66] into”
BodyNet: Volumetric Inference of 3D Human Body Shapes (2018, neutral): “Recent work explores voxel [20–23], octree [24–27], point cloud [28–30], and surface [31] representations for modeling generic 3D objects.”
Deep Global Registration (2020, methodology): “We follow the train/test split and the standard procedure to generate pairs with at least 30% overlap for training and testing [12, 11, 9].”
Image Matching from Handcrafted to Deep Features: A Survey (2020, background): “Deng et al. (2018) first presented an order-free network on the basis of PointNet (Qi et al. 2017a).”
Feature-Metric Registration: A Fast Semi-Supervised Approach for Robust Point Cloud Registration Without Correspondences (2020, methodology): “Different to the classical optimization-based registration methods, numerous recent registration methods [27, 4, 9] are tried to use the deep neural network to learn a robust descriptor.”
A comprehensive survey on point cloud registration (2021, methodology): “[19] proposes a local PPF feature by using the distribution of neighbour points and then input into the network for deep feature learning.”
The Perfect Match: 3D Point Cloud Matching With Smoothed Densities (2018, methodology): “Since achieving the rotation invariance in practice is non-trivial, several recent works [49, 5, 45] choose to ignore it and thus do not generalize to rigidly transformed scenes [4].”
PointDSC: Robust Point Cloud Registration using Deep Spatial Consistency (2021, methodology): “Recently, learningbased algorithms have been proposed to replace the individual components in the classical registration pipeline, including keypoint detection [4, 38, 32] and feature description [21, 22, 23, 51, 4, 18, 27, 30, 2].”
D3Feat: Joint Learning of Dense Detection and Description of 3D Local Features (2020, background): “[5, 4] build their network upon PointNet to directly consume unordered point sets.”
3DFeat-Net: Weakly Supervised Local 3D Features for Point Cloud Registration (2018, background): “…approaches of the local 3D feature detectors and descriptors are heavily inﬂuenced by the widely studied 2D local features methods [17,2], and can be broadly categorized into handcrafted [35,5,26,25,13,30,28] and learning approaches [34,8,15,27,6] - i.e. pre-and post-deep learning approaches.”
RPM-Net: Robust Point Matching Using Learned Features (2020, methodology): “Voxelization re-sults in a loss of quality, so later works such as PPFNet [7] uses a PointNet [23, 24] architecture to learn features directly from raw point clouds.”
Deep Learning for LiDAR Point Clouds in Autonomous Driving: A Review (2020, neutral): “Although multi-scale feature fusion approaches [142–144] have”