Digital Images
4 filmsComing soonWhat a picture is to a machine before anything is recognised in it: a grid of numbers of a chosen size, in three colour layers, and the one picture made for machines to read, the QR code.
Your phone unlocks on your face, your photos app finds 'beach', your car reads the lane and your camera reads a QR code. Every one of them starts from the same thing: a grid of numbers. One idea at a time, with the pixels moving, in plain words.
What a picture is to a machine before anything is recognised in it: a grid of numbers of a chosen size, in three colour layers, and the one picture made for machines to read, the QR code.
How a machine finds structure in a grid of numbers: edges as jumps between neighbours, convolution as one small window slid everywhere, corners matched between two pictures, the layers a network learns, and naming a whole picture.
The recognition jobs inside the products people use daily: finding every thing in a frame, marking its exact pixels, reading text, finding things that look alike, and matching a face, each built from one design idea.
Vision systems built for the world: depth from one camera, an object followed across frames, a car's cameras fused into one map, a medical scan read in a real clinic, the inputs that fool a model, and the models that turn pictures into words.