2D Images Obstacle Labeling Case Study

In the automatic driving scenario, the moving vehicle needs to analyze the position relationship with other vehicles in the current lane and make appropriate control decisions to avoid accidents.

For example, if there is a vehicle in a close distance in front, or vehicles on both sides are trying to cut into the lane, this vehicle needs to slow down in time.

In order to obtain the accurate distance of other vehicles relative to this vehicle, it is necessary to obtain the accurate location information of other vehicles, and then estimate the distance between vehicles.

The artificial neural network model is used to detect the rectangular box containing the car, and the distance between the ego car and the neighbor one can be further estimated according to the position. For example, in vehicle detection, the horizontal distance between two vehicles can be estimated according to the abscissa of the midpoint of the bottom border of the box.

Let’s share a visual obstacle detection annotation case.

1. Obstacles Categories

Car、Bus、Truck、Rider、Tricycle、Bicycle、Pedestrian

2. Labeling Requirements

Labeling type: 2D boxing

For Cars 、Bus、 Trucks, and moving vehicles, the direction is needed.

Obstacle size: more than 15*15 pixels

Labeling level: down to 5 pixels.

3. Obstacles Categories and Labeling Attributes

4. Car Labeling Details

A. Basic Rules

- For truncation and crowding obstacles, the whole part should be evaluated

- Accessories like Taxi top lights and rearview mirrors don’t need to be labeled

B. Occlusion Types( Occluded by the background)

- Not occluded

- Partial occluded (percentage)

- Highly occluded (percentage)

C. Crowding Types(Occluded by other obstacles)

- Not crowding

- Partial crowding (percentage)

- Highly crowding (percentage)

D. Truncation Types

- Not Truncation

- Truncation

E. Direction

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)

Why not have a try?