Machine vision lets a computer look at images, live or recorded and figure out what’s in them. While this means party tricks like telling dogs from cats, or distinguishing hot dogs from, well, any other object, it can also be used to free robots up to do more than just repetitive tasks, meaning a wider range of applications and more flexible approaches.

SEE: Inside Amazon’s clickworker platform: How half a million people are being paid pennies to train AI (PDF download) (TechRepublic)

Here are the top five things to know about machine vision:

1. It’s a big market. BCC Research pegs the global market for machine vision system components at $19.0 billion for 2016 and an estimated $30.8 billion by 2021

2. One camera is all you need to get some benefit. If you need a robot to sort items, let’s say cookies, one camera in a fixed position can view the cookies as they come out of an oven on a conveyor belt and sort the different kinds into bins.

3. Machine vision does quality control! Robots can also learn how to tell shapes from shadows and look for defects based on an image of how an object is supposed to look. This shape from shading process can evaluate as a fast as two parts per second in some cases.

SEE: Understanding the differences between AI, machine learning, and deep learning

4. A type of AI called neural networks is increasing machine vision capability. Neural networks can help segment images and classify and detect objects better.

5. Robots can be used to determine the position of objects using stereo vision. In other systems, time-of-flight sensors can scan objects to make 3D models for multiple purposes. One was able to read the black codes of raised rubber on black tires this way.

For more on machine learning and artificial intelligence, check out these articles from TechRepublic: