Computer vision: how robots see
BEGINNER · 6 MIN READ
A camera gives a robot a grid of numbers. Computer vision turns those numbers into "that is a cup".
A camera does not see a cup. It records a grid of tiny coloured dots, each one just a number. Computer vision is the set of methods that turns that grid into meaning: edges, shapes, objects, distances.
Modern robots mostly use learned models for this. They find objects in the picture, draw a box around each, and name it. Some also work out how far away things are, using two cameras or a depth sensor.
A camera does not see a shape. It records numbers. Each square below is one number (0 is black, 255 is white). The robot has one simple rule: anything darker than 100 is an object.
Robot says: Object found (32 squares darker than 100)
Now drag brightness down to 40%. The table gets darker than the rule, so the whole picture looks like an object. Same scene, different light, wrong answer. That is why real robots learn what objects look like in many lights, instead of using one fixed rule.
A fruit-picking robot finds the ripe strawberries in a photo, works out where each one is in 3D, and moves its gripper to that spot.
Your robot can find, count, follow and pick up things without being told exactly where they are.
The robot is blind to anything that is not in a fixed, known place. A shifted tray means a missed pick.
Hold your phone camera on a table of objects. Imagine writing the numbers a robot gets. Which would be hardest to tell apart, and why?
Common questions
Do robots need expensive cameras?
Often not. A cheap camera with good lighting beats an expensive one in poor light.
Why does vision fail in bad light?
Dim or glaring light changes the numbers the camera records, so objects look different from what the model learned.
Keep learning
Get your personal invite link to share R2BOT with friends.
Teachers and bloggers: paste the embed code into your page to show this demo for free.
Plain English. Free. Unsubscribe in one click.