- 2D Bounding Boxes: The rectangular boxes are made use of to identify objects, label and to categorize. You need to draw the boxes manually around objects in motion across many frames. The box is placed closer to the edge of the object and labelled correctly for classes and attributes. It helps in the depiction of the object and its movement in all frames with accuracy.
- 3D Bounding Boxes: This method is used to obtain a more realistic and accurate depiction of an object and its interaction with its surroundings. If you want to detect common or particular classes of items, the 3D bounding boxes method works the best.
- Polygons: The polygon method is mostly used when the 3D or 2D bounding boxes are inadequate to rightly depict an item in motion. Annotators need to form lines and hence they place dots across the outer border of the object for precise annotation.
- Key Point or Landmark: The landmark annotation method means to generate dots on the image and link them to create a skeleton of the object of interest. It helps in identifying the smallest of items, shapes and postures.
- Lines and Splines: The method helps robots to learn and identify borders and lanes in the autonomous driving field. The lines are drawn by the annotators between areas that help the AI program to identify across frames.