Agentic data science with the Data Agent Kit
cloud.google.com·8h ago
AI Research
arXiv CS.AI·Aug 31·5 min read
Depth-Aware Pothole Detection Using YOLO and RT-DETR at the Edge
TL;DR
A depth-aware pothole detection framework was developed using YOLO and RT-DETR models, achieving high accuracy in detecting potholes and estimating their depth. YOLOv8nSeg outperformed others with a mean Average Precision (mAP@50) of 0.9556 and a depth estimation accuracy of 2.96 cm.
✦ Why It Matters
Implement depth-aware pothole detection in urban maintenance systems to enhance road safety and reduce repair costs.
Key Takeaways
- Utilize YOLOv8nSeg for high-accuracy pothole detection.
- Incorporate RGB-D sensors for depth measurement in road monitoring.
- Apply RANSAC for correcting depth map errors.
- Consider structural biases in depth estimation when using bounding boxes.
How It Works
The framework combines RGB images with depth data from RGB-D sensors to enhance pothole detection. It employs various YOLO and RT-DETR models to analyze the data, with a focus on depth estimation accuracy.
The Dseg algorithm refines depth measurements, correcting for camera tilt and ensuring precise statistics.