What are the methods of data annotation?

The methods of data annotation include rectangular box annotation, polygon annotation, key point annotation, 3D point cloud annotation, 3D cube annotation and so on.

1, rectangular box labeling

Rectangular frame labeling is a simple way to label target objects with target detection frame, which is often used to label people, cars and things under automatic driving.

2. Polygon labeling

Polygon labeling refers to labeling irregular target objects with polygonal frames in static pictures. Compared with rectangular box labeling, polygon labeling can frame the target more accurately and be more targeted for irregular objects.

3. Key point marking

The biggest application of key point labeling template is to label the key points of human face. By marking the key points in different directions, we can judge the function of people in the image.

4.3D point cloud labeling

Three-dimensional point cloud labeling refers to frame selection and labeling using data collected by lidar, which is used to train artificial intelligence models such as computer vision and unmanned driving.

5, three-dimensional cube labeling

Different from point cloud labeling, 3D cube labeling is still based on 2D plane images. The labeling machine frames the edges of three-dimensional objects to obtain vanishing points and measure the relative distance between objects.