DeFlow: Decoder of Scene Flow Network in Autonomous Driving (2024-01-29T00:00:00.000000Z)

TL;DR

This paper introduces DeFlow which enables a transition from voxel-based features to point features using Gated Recurrent Unit (GRU) refinement, and formulate a novel loss function that accounts for the data imbalance between static and dynamic points.

Abstract

Scene flow estimation determines a scene’s 3D motion field, by predicting the motion of points in the scene, especially for aiding tasks in autonomous driving. Many networks with large-scale point clouds as input use voxelization to create a pseudo-image for real-time running. However, the voxelization process often results in the loss of point-specific features. This gives rise to a challenge in recovering those features for scene flow tasks. Our paper introduces DeFlow which enables a transition from voxel-based features to point features using Gated Recurrent Unit (GRU) refinement. To further enhance scene flow estimation performance, we formulate a novel loss function that accounts for the data imbalance between static and dynamic points. Evaluations on the Argoverse 2 scene flow task reveal that DeFlow achieves state-of-the-art results on large-scale point cloud data, demonstrating that our network has better performance and efficiency compared to others. The code is available at https://github.com/KTH-RPL/deflow.

Authors

Ruoyu Geng

3 papers

P. Jensfelt

2 papers

Qingwen Zhang

1 papers

TL;DR

Abstract

Authors

References36 items

DUFOMap: Efficient Dynamic Awareness Mapping

A Dynamic Points Removal Benchmark in Point Cloud Maps

3D Point-Voxel Correlation Fields for Scene Flow Estimation

Dynablox: Real-Time Detection of Diverse Dynamic Objects in Complex Environments

Fast Neural Scene Flow

Argoverse 2: Next Generation Datasets for Self-Driving Perception and Forecasting

SCOOP: Self-Supervised Correspondence and Optimization-Based Scene Flow

Motion Inspired Unsupervised Perception and Prediction in Autonomous Driving

GMA3D: Local-Global Attention Learning to Estimate Occluded Motions of Scene Flow

Patchwork++: Fast and Robust Ground Segmentation Solving Partial Under-Segmentation Using 3D Point Cloud

Bi-PointFlowNet: Bidirectional Learning for Point Cloud Based Scene Flow Estimation

Open-world Semantic Segmentation for LIDAR Point Clouds

Point-to-Voxel Knowledge Distillation for LiDAR Semantic Segmentation

CRAFT: Cross-Attentional Flow Transformer for Robust Optical Flow

GMFlow: Learning Optical Flow via Global Matching

Neural Scene Flow Prior

Separable Flow: Learning Motion Cost Volumes for Optical Flow Estimation

Self-Supervised Pillar Motion Learning for Autonomous Driving

Scalable Scene Flow From Point Clouds in the Real World

PV-RAFT: Point-Voxel Correlation Fields for Scene Flow Estimation of Point Clouds

RAFT: Recurrent All-Pairs Field Transforms for Optical Flow

Scalability in Perception for Autonomous Driving: Waymo Open Dataset

End-to-End Multi-View Fusion for 3D Object Detection in LiDAR Point Clouds

Iterative Residual Refinement for Joint Optical Flow and Occlusion Estimation

PointPillars: Fast Encoders for Object Detection From Point Clouds

PWC-Net: CNNs for Optical Flow Using Pyramid, Warping, and Cost Volume

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

ShapeNet: An Information-Rich 3D Model Repository

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

Multi-Scale Context Aggregation by Dilated Convolutions

Object scene flow for autonomous vehicles

FlowNet: Learning Optical Flow with Convolutional Networks

Fast segmentation of 3D point clouds for ground vehicles

ZeroFlow: Fast Zero Label Scene Flow via Distillation

TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE

“Argoverse 2 scene flow online leaderboard

Field of Study

Journal Information

Name

Page

Venue Information

Name

Type

URL

Alternate Names