There are three types of computer vision models:

  • Classification
  • Detection
  • Segmentation

1. Classification Model

Classification functions by determining a given “class” that an image, or an object in an image, belongs to, from a simple yes/no to thousands of classes. There is a type of “probability” by class, so that the highest probability gets to be the determined class. However, you can also see the top 5 predictions as well.

2. Detection Model

The Detection model gets into determining the objects that appear at different places in an image, and it often draws bounding boxes around the detected objects.

In addition, it does have some form of classification that determines the class of an object in a given bounding box. The bounding boxes have a confidence threshold so you can throw out low-confidence detections.

3. The Segmentation Model

The Segmentation does classify sections of an image by classifying each and every pixel.

The networks are often post-processed in some way to avoid phantom classes. Within the segmentation, there are the subsets of semantic segmentation and instance segmentation.

For instance, the first on all instances of a class are considered as one, while the second is considered as separate instances of a class as separate objects.


If you have any question or comment, do not hesitate to ask us.

Quote: The moon looks upon many night flowers; the night flowers see but one moon. – Jean Ingelow