Think together · Start with a story
A girl zooms in on a photo and sees that it is made of many tiny pixels. The robot processes this kind of information too, in a form it can compute.
Break down the process
- PixelsValues for color and brightness
- RepresentationEncoded for the model
- ModelCalculates visual patterns
- OutputA category or another result
AI first represents images with values such as pixels, then uses a model to extract and combine information.
Different models may organize input using local regions or image patches. A person sees a whole kitten; a model calculates a representation that is useful for its task.
Image size, cropping, and clarity affect what information is available.
Your turn to try
Why might recognition get worse when the same object is heavily blocked from view?
Ready? Open to see the explanation
There are fewer clues to use, so the model may not be able to distinguish between possibilities. It needs a more complete view.
One idea to remember
AI first represents images with values such as pixels, then uses a model to extract and combine information.
Try covering the picture above and explaining each step to your family. If one step is hard to explain, revisit that card.