Title
DeepPoint3D: Learning Discriminative Local Descriptors using Deep Metric Learning on 3D Point Clouds.
Abstract
•We learn a 3D descriptor encoding the local neighbourhood of a 3D point by directly processing unstructured 3D point clouds.•We propose and evaluate three siamese-style deep architectures to learn such encodings from raw 3D point clouds.•We propose a novel multi margin contrastive loss to enhance learning by exploiting extent of variation in the training data.•We outperform the state-of-the-art methods by ∼3−5% on various metrics for both high and low resolution data.•We demonstrate the generalization ability of our method by achieving state-of-the-art results on various applications.
Year
DOI
Venue
2019
10.1016/j.patrec.2019.02.027
Pattern Recognition Letters
Keywords
Field
DocType
41A05,41A10,65D05,65D17
Pattern recognition,Convolutional neural network,Permutation,Image based,RGB color model,Invariant (mathematics),Artificial intelligence,Intermediate language,Point cloud,Discriminative model,Mathematics
Journal
Volume
ISSN
Citations 
127
0167-8655
1
PageRank 
References 
Authors
0.37
0
2
Name
Order
Citations
PageRank
Siddharth Srivastava195.89
Brejesh Lall28543.42