Real-Time 3D Object Detection Using InnovizOne LiDAR and Low-Power Hailo-8 AI Accelerator

Itay Krispin-Avraham,Roy Orfaig,Ben-Zion Bobrovsky

2024-12-07

Abstract:Object detection is a significant field in autonomous driving. Popular sensors for this task include cameras and LiDAR sensors. LiDAR sensors offer several advantages, such as insensitivity to light changes, like in a dark setting and the ability to provide 3D information in the form of point clouds, which include the ranges of objects. However, 3D detection methods, such as PointPillars, typically require high-power hardware. Additionally, most common spinning LiDARs are sparse and may not achieve the desired quality of object detection in front of the car. In this paper, we present the feasibility of performing real-time 3D object detection of cars using 3D point clouds from a LiDAR sensor, processed and deployed on a low-power Hailo-8 AI accelerator. The LiDAR sensor used in this study is the InnovizOne sensor, which captures objects in higher quality compared to spinning LiDAR techniques, especially for distant objects. We successfully achieved real-time inference at a rate of approximately 5Hz with a high accuracy of 0.91% F1 score, with only -0.2% degradation compared to running the same model on an NVIDIA GeForce RTX 2080 Ti. This work demonstrates that effective real-time 3D object detection can be achieved on low-cost, low-power hardware, representing a significant step towards more accessible autonomous driving technologies. The source code and the pre-trained models are available at <a class="link-external link-https" href="https://github.com/AIROTAU/" rel="external noopener nofollow">this https URL</a> PointPillarsHailoInnoviz/tree/main

Computer Vision and Pattern Recognition,Artificial Intelligence

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: How to achieve efficient real - time 3D object detection on low - cost, low - power - consumption hardware, so as to promote the popularization and application of autonomous driving technology.

Specifically, the paper focuses on the following points:

1. **High - power - consumption problem**: Traditional 3D object detection methods (such as PointPillars) usually require high - performance, high - power - consumption hardware (such as NVIDIA GeForce RTX 2080 Ti), which limits their scalability and cost - effectiveness in practical applications.

2. **Limitations of LiDAR sensors**: Common rotating LiDAR sensors may not be able to provide sufficient precision in some cases (such as long - distance object detection), affecting the quality of object detection.

3. **Requirement for real - time processing**: In order to realize the safe navigation and decision - making of autonomous driving vehicles, 3D object detection must be able to operate efficiently in a real - time environment.

To solve these problems, the paper proposes a solution based on the InnovizOne LiDAR sensor and the Hailo - 8 AI accelerator. By optimizing the PointPillars model and deploying it on the low - power - consumption Hailo - 8 chip, the research team has successfully achieved a real - time inference speed of about 5Hz and maintained a relatively high detection accuracy (the F1 score reaches 0.91%), with only a - 0.2% decrease compared to the performance on a high - performance GPU.

In addition, the paper also explores how to apply this low - power - consumption solution to a wider range of autonomous systems, thereby reducing the cost and power - consumption requirements of autonomous driving technology and making it more popular and practical.