Case Study of Traffic Light Data Annotation
“Run a Red Light”
A self-driving company conducted an unmanned vehicle test. During the test, the unmanned vehicle accurately recognized the red light, but the system ignored the message, and the vehicle continued to drive through the intersection. No measures were taken, and fortunately, no casualties were caused. In the incident, the camera recognized the red signal light, possibly because the processor considered that the red light was the taillight of the car in front.
External factors such as weather, tail lights of surrounding vehicles, etc., may interfere with automatic driving technology. Being able to understand traffic signals and make corresponding responses play an important role in self-driving perception and decision-made sectors. The premise is that the training data of traffic lights in different scenarios, weather, and categories are available.
Let’s have a look at a traffic light data labeling case.
Basic Information About the Annotation Project:
Task Introduction
For the ego vehicles running on the road, they rely on the camera sensor to detect and identify the traffic lights, it is necessary to label the location and status of the traffic lights for large-scale data training.
All traffic lights, portable traffic lights, timers, road traffic lights, flashing warning lights, non-vehicles, and crossing traffic lights in the front view should be labeled with 2D bounding boxes.
Make sure to only label the front and lateral lights, and don’t label backward ones.
Labeling Categories
Box the traffic sign with different categories. The specific categories are shown as follows:
Project Introduction
1 Project type: Image annotation
2 The number of images: more than 1000 K
3 Delivery time: 2 months
4 Delivery requirements: Labeling down to 5 Pixels and the accuracy rate should be no less than 99%.
5 Labeling action: bounding box+ classification
Basic Labeling Principles
A traffic sign is usually composed of multiple child lights which can even reach four. Each light presents a state, including color and shape. Therefore, we need to give the information about both the position and state, which include:
1 Each traffic sign should be marked with a 2D bounding box
2 The status information is tagged with a multi-category description, presenting the status of working lights, from top to bottom and from left to right.
You Configure and ByteBridge Annotates MANUALLY
Output
JSON Output
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)
