SqueezeSegV2: Improved Model Structure and Unsupervised Domain Adaptation for Road-Object Segmentation from a LiDAR Point Cloud

Abstract

Earlier work demonstrates the promise of deep-learning-based approaches for point cloud segmentation; however, these approaches need to be improved to be practically useful. To this end, we introduce a new model SqueezeSegV2 that is more robust to dropout noise in LiDAR point clouds. With improved model structure, training loss, batch normalization and additional input channel, SqueezeSegV2 achieves significant accuracy improvement when trained on real data. Training models for point cloud segmentation requires large amounts of labeled point-cloud data, which is expensive to obtain. To sidestep the cost of collection and annotation, simulators such as GTA-V can be used to create unlimited amounts of labeled, synthetic data. However, due to domain shift, models trained on synthetic data often do not generalize well to the real world. We address this problem with a domain-adaptation training pipeline consisting of three major components: 1) learned intensity rendering, 2) geodesic correlation alignment, and 3) progressive domain calibration. When trained on real data, our new model exhibits segmentation accuracy improvements of 6.0-8.6% over the original SqueezeSeg. When training our new model on synthetic data using the proposed domain adaptation pipeline, we nearly double test accuracy on real-world data, from 29.0% to 57.4%. Our source code and synthetic dataset will be open-sourced.

arXiv commentBichen Wu, Xuanyu Zhou, and Sicheng Zhao contributed equally to this paper

References in corpus (6)

• Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
• Learning Transferable Features with Deep Adaptation Networks
• Curriculum Domain Adaptation for Semantic Segmentation of Urban Scenes
• Domain Adaptation for Visual Applications: A Comprehensive Survey
• PointSeg: Real-Time Semantic Segmentation Based on 3D LiDAR Point Cloud
• Minimal-Entropy Correlation Alignment for Unsupervised Deep Domain Adaptation

Cited by in corpus (12)

• Augmented LiDAR Simulator for Autonomous Driving
• FBNet: Hardware-Aware Efficient ConvNet Design via Differentiable Neural Architecture Search
• From SLAM to Situational Awareness: Challenges and Survey
• FG-Net: Fast Large-Scale LiDAR Point Clouds Understanding Network Leveraging Correlated Feature Mining and Geometric-Aware Modelling
• SqueezeWave: Extremely Lightweight Vocoders for On-device Speech Synthesis
• RIU-Net: Embarrassingly simple semantic segmentation of 3D LiDAR point cloud
• Robust Multi-Modality Multi-Object Tracking
• PASS3D: Precise and Accelerated Semantic Segmentation for 3D Point Cloud
• Efficient Deep Neural Networks
• LATTE: Accelerating LiDAR Point Cloud Annotation via Sensor Fusion, One-Click Annotation, and Tracking
• FusionMapping: Learning Depth Prediction with Monocular Images and 2D Laser Scans
• LiDAR-based Recurrent 3D Semantic Segmentation with Temporal Memory Alignment