Occlusion Guided Self-supervised Scene Flow Estimation on 3D Point Clouds (2021-04-10T00:00:00.000000Z)

TL;DR

This work presents a new self-supervised training method and an architecture for the 3D scene flow estimation under occlusions that outperforms traditional architectures by a large margin for occluded and non-occluded scenarios.

Abstract

Understanding the flow in 3D space of sparsely sampled points between two consecutive time frames is the core stone of modern geometric-driven systems such as VR/AR, Robotics, and Autonomous driving. The lack of real, non-simulated, labeled data for this task emphasizes the importance of self- or un-supervised deep architectures. This work presents a new self-supervised training method and an architecture for the 3D scene flow estimation under occlusions. Here we show that smart multi-layer fusion between flow prediction and occlusion detection outperforms traditional architectures by a large margin for occluded and non-occluded scenarios. We report state-of-the-art results on Flyingthings3D and KITTI datasets for both the supervised and self-supervised training.1

Authors

D. Raviv

4 papers

Bojun Ouyang

1 papers

TL;DR

Abstract

Authors

References53 items

Occlusion Guided Scene Flow Estimation on 3D Point Clouds

FlowStep3D: Model Unrolling for Self-Supervised Scene Flow Estimation

Adversarial Self-Supervised Scene Flow Estimation

Self-Supervised Learning of Non-Rigid Residual Flow and Ego-Motion

DeepLiDARFlow: A Deep Learning Architecture For Scene Flow Estimation Using Monocular Camera and Sparse LiDAR

FLOT: Scene Flow on Point Clouds Guided by Optimal Transport

Self-Supervised Monocular Scene Flow Estimation

Learning by Analogy: Reliable Supervision From Transformations for Unsupervised Optical Flow Estimation

MaskFlownet: Asymmetric Feature Matching With Learnable Occlusion Mask

Just Go With the Flow: Self-Supervised Scene Flow Estimation

Point-Voxel CNN for Efficient 3D Deep Learning

HPLFlowNet: Hierarchical Permutohedral Lattice FlowNet for Scene Flow Estimation on Large-Scale Point Clouds

SelFlow: Self-Supervised Learning of Optical Flow

PWOC-3D: Deep Occlusion-Aware End-to-End Scene Flow Estimation

Iterative Residual Refinement for Joint Optical Flow and Occlusion Estimation

DDFlow: Learning Optical Flow with Unlabeled Data Distillation

PointConv: Deep Convolutional Networks on 3D Point Clouds

Models Matter, So Does Training: An Empirical Study of CNNs for Optical Flow Estimation

Unsupervised Learning of Multi-Frame Optical Flow with Occlusions

Occlusions, Motion and Depth Boundaries with a Generic Network for Disparity, Optical Flow or Scene Flow Estimation

PointFlowNet: Learning Representations for Rigid Motion Estimation From Point Clouds

FlowNet3D: Learning Scene Flow in 3D Point Clouds

Deep Parametric Continuous Convolutional Neural Networks

Occlusion Aware Unsupervised Learning of Optical Flow

UnFlow: Unsupervised Learning of Optical Flow with a Bidirectional Census Loss

Multiframe Scene Flow with Piecewise Rigid Motion

MirrorFlow: Exploiting Symmetries in Joint Optical Flow and Occlusion Estimation

PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space

A learning approach for real-time temporal scene flow estimation from LIDAR data

PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation

Rigid scene flow for 3D LiDAR scans

Volumetric and Multi-view CNNs for Object Classification on 3D Data

A Large Dataset to Train Convolutional Networks for Disparity, Optical Flow, and Scene Flow Estimation

Blind video temporal consistency

Joint 3d Estimation of Vehicles and Scene Flow

Object scene flow for autonomous vehicles

FlowNet: Learning Optical Flow with Convolutional Networks

3D ShapeNets: A deep representation for volumetric shapes

Affine Invariant Geometry for Non-rigid Shapes

Scale Invariant Geometry for Nonrigid Shapes

Kinecting the dots: Particle based scene flow from depth sensors

Hierarchical Matching of Non-rigid Shapes

Affine-invariant diffusion geometry for the analysis of deformable 3D shapes

Multi-view Scene Flow Estimation: A View Centered Variational Approach

A Variational Method for Scene Flow Estimation from Stereo Sequences

Optimal Step Nonrigid ICP Algorithms for Surface Registration

Three-dimensional scene flow

A Method for Registration of 3-D Shapes

Object modeling by registration of multiple range images

PointPWC-Net: Cost Volume on Point Clouds for (Self-)Supervised Scene Flow Estimation

Graph Attention Convolution for Point Cloud Semantic Segmentation

VoxNet: Reducing Latency in High Data Rate Applications

upsam-pled ﬂow from level l + 1

Field of Study

Journal Information

Name

Page

Venue Information

Name

Type

URL

Alternate Names