Road Arrow Polygon Annotation Case Study
We often see various arrows on the road. These arrows are road traffic markers with a guiding function, indicating that the road ahead can only be driven according to the guiding arrows on the road surface. They mainly provide guiding information to traffic participants. The indicator type generally gives pedestrians and drivers exact road traffic information to make the road traffic smooth and safe.
How do People usually Drive?
Observation: Observe Road conditions, such as vehicle distance, pedestrians, traffic signs, etc.;
Analysis: analyze whether to accelerate, decelerate, or stop based on experience;
Make decisions: Act based on actual conditions, such as pressing the accelerator or brake.
The process of autonomous driving is no different from that of a human driver:
Perception: collecting images and other data through the camera, radar, and other hardware is equivalent to the observation of human eyes;
Analysis: real-time analysis of images through deep learning models to make decisions for the next step;
Making decisions: in order for driverless vehicles to be able to drive on the road safely and without obstacles, we need to allow vehicles to autonomously identify various types of road traffic markers. The scenarios for data annotation usually include lane changing and overtaking, passing intersections, and unprotected left and right turns without traffic light control.
Let’s look at a case of road arrow polygon annotation.
Labeling Types
Key points annotation: label the road arrows with key points. Each type of arrow has a fixed number of points. The points are labeled counterclockwise and connected in order.
The arrow names and the number of key points are as follows.
- arrow_straight：7 points.
- arrow_left：9 points.
- arrow_right：9 points.
- arrow_merge_left：9 points.
- arrow_merge_right：9 points.
- arrow_left_right：15 points.
- arrow_straight_left：21 points
- arrow_straight_right：21 points.
- arrow_uturn：13 points.
- arrow_ruturn：13 points
- arrow_uturn_straight：18 points
- arrow_ruturn_straight：18 points
- arrow_uturn_left：20 points.
- arrow_ruturn_right：20 points
- arrow_straight_left_right：14 points
Labeling Requirements
- Each point should be labeled counterclockwise in strict accordance with its corresponding position
- If it is not clear what type of arrow it is due to occlusion, or if part of it is hard to tell, don’t label it
- If the position of one point cannot be accurately estimated, an additional label should be added for this point as the imagined point
- There shall be no missing or wrong labeling
End
Outsource your data labeling tasks to ByteBridge, you can get high-quality ML training datasets cheaper and faster!
- Free Trial Without Credit Card: you can get your sample result in a fast turnaround, check the output, and give feedback directly to our project manager.
- 100% Human Validated
- Transparent & Standard Pricing: clear pricing is available(labor cost included)
