UNet-Based Keypoint Regression for 3D Cone Localization in Autonomous Racing

TL;DR

This paper introduces a UNet-based neural network for precise 3D cone localization in autonomous racing, demonstrating significant accuracy improvements and effective integration into the perception pipeline for real-time navigation.

Contribution

The paper presents a novel UNet-based keypoint detection method trained on a large custom dataset, improving 3D cone localization accuracy for autonomous racing.

Findings

Significant accuracy improvements over traditional methods

Effective integration into autonomous perception pipeline

High performance across all evaluation metrics

Abstract

Accurate cone localization in 3D space is essential in autonomous racing for precise navigation around the track. Approaches that rely on traditional computer vision algorithms are sensitive to environmental variations, and neural networks are often trained on limited data and are infeasible to run in real time. We present a UNet-based neural network for keypoint detection on cones, leveraging the largest custom-labeled dataset we have assembled. Our approach enables accurate cone position estimation and the potential for color prediction. Our model achieves substantial improvements in keypoint accuracy over conventional methods. Furthermore, we leverage our predicted keypoints in the perception pipeline and evaluate the end-to-end autonomous system. Our results show high-quality performance across all metrics, highlighting the effectiveness of this approach and its potential for adoption in competitive autonomous racing systems.