How computers see pictures
Big idea: to a computer, every picture is a grid of numbers. Seeing a picture is not the same as understanding it.
The story in short
Kabir chachu zoomed into a photo of a guava (amrood) leaf on his laptop, closer and closer, until it became tiny coloured squares, each with three numbers like “R 220 · G 40 · B 40”. The children drew a number rangoli to see how a picture can be made of numbers. Then the app looked at a photo of Mitthu hiding among the leaves and called him a leaf, but only “61% sure”, not “94% sure” like before.
The idea
- Pixels: the tiny single-colour squares that make up a digital picture.
- RGB values: three numbers from 0 to 255 for how much red, green and blue light to mix.
- Edge detection: spotting where the numbers suddenly change, like where green leaf meets blue sky.
- Layers of features: edges, then shapes, then parts, then “leaf”: simple patterns combined step by step.
- Computer vision: teaching computers to find meaning in pictures.
- Misclassification: putting something in the wrong category (a green parrot as a leaf).
- Confidence score: how sure the AI is. A low number means: check!
Where you see it
Face unlock, Google Lens, scanning a UPI QR code, number-plate cameras at toll plazas: all of them turn pictures into numbers first.
Remember: step back and the parts make a whole, but a computer only sees numbers until it has learned what they mean.
Check yourself
A pixel shows R 255 · G 255 · B 255. What colour is it?
What does edge detection look for in a picture?
A photo app labels your green parrot "leaf, 61% sure". What does the 61% tell you?
More for you: 🧒 Kids · 🧑🏫 Parents and teachers