Deep Learning for Semantic Segmentation of 3D Ultrasound Data

TL;DR

This paper presents a novel 3D ultrasound-based semantic segmentation framework using a 3D U-Net architecture, demonstrating robust performance and highlighting ultrasound sensing as a promising modality for autonomous systems in challenging environments.

Contribution

Introduces a new 3D ultrasound sensor system and a deep learning framework for volumetric segmentation, expanding perception options for autonomous vehicles in harsh conditions.

Findings

Robust segmentation performance with Calyo Pulse sensors

Potential improvements with larger datasets and refined ground truth

Highlights 3D ultrasound as a promising modality for autonomous perception

Abstract

Developing cost-efficient and reliable perception systems remains a central challenge for automated vehicles. LiDAR and camera-based systems dominate, yet they present trade-offs in cost, robustness and performance under adverse conditions. This work introduces a novel framework for learning-based 3D semantic segmentation using Calyo Pulse, a modular, solid-state 3D ultrasound sensor system for use in harsh and cluttered environments. A 3D U-Net architecture is introduced and trained on the spatial ultrasound data for volumetric segmentation. Results demonstrate robust segmentation performance from Calyo Pulse sensors, with potential for further improvement through larger datasets, refined ground truth, and weighted loss functions. Importantly, this study highlights 3D ultrasound sensing as a promising complementary modality for reliable autonomy.