There are a handful of different ways that computers can look at images, and one of the options is “stuff” versus “things.” Instance segmentation can be used to find “things:” each individual person, car, tree, etc in an image.
Yes, this is the object detection page but the image is more or less panoptic segmentation, which detects both the objects and what pixels make them up. It’s like this plus semantic segmentation.
Use cases
Instance segmentation is great at answering the question of “is XXXX in this photo?” You can think of it as an image classification model with a little bit more nuance.
You might want to filter thousands of photos to only show the ones that include a car or an elephant, or count the number of people in the image of a crowd.
Instance segmentation also shows you where in the image something is, which may or may not be useful.
Try it out
The examples are not as simple as I’d like, but we’ll go with this one for now. Be sure to scroll down to click the examples!