Deep Learning for Semantic Segmentation of 3D Ultrasound Data
It addresses the need for cost-efficient and reliable perception in automated vehicles, offering a complementary modality to LiDAR and cameras, though it appears incremental as it adapts existing methods to new sensor data.
This work tackled the problem of 3D semantic segmentation for automated vehicles by introducing a learning-based framework using Calyo Pulse ultrasound sensors, achieving robust segmentation performance with potential for further improvement.
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.