One-Stage Brake Light Status Detection Based on YOLOv8

Area of Science:

• Computer Vision
• Artificial Intelligence
• Automotive Engineering

Background:

• Advanced driver assistance systems (ADAS) and autonomous driving face challenges in achieving Level 3 automation.
• Level 3 automation requires enhanced safety and interpretability, necessitating reliable vehicle cue detection.
• Brake light status is a critical visual cue for human drivers and autonomous systems.

Purpose of the Study:

• To develop a novel method for detecting driving vehicles and their brake light status.
• To improve the safety, interpretability, and comfort of advanced driver assistance systems (ADAS) and autonomous driving technologies.
• To provide a publicly available dataset for brake light status detection research.

Main Methods:

• A fast and accurate one-stage brake light status detection network based on YOLOv8 was developed.
• Transfer learning was applied using a custom dataset of over 11,000 forward images with manual annotations.
• The method was evaluated for detection accuracy and inference time on an edge device (Jetson Nano).

Main Results:

• The proposed YOLOv8 network accurately detects vehicles and their brake light status.
• High detection performance was achieved with a mean average precision (mAP50) between 0.766 and 0.793.
• Short inference time of 133.30 ms was recorded on the Jetson Nano edge device.

Conclusions:

• The proposed method achieves high accuracy and fast inference for brake light status detection.
• This contributes valuable input information for ADAS and autonomous driving, enhancing safety and interpretability.
• The publicly available dataset supports further research in this critical area of automotive technology.