Lesson 7.10ModelCode~1 h
Convolutional networks
Slide a small learned filter across an image: edges, textures, shapes — how machines learned to see.
Module 7 · 0/11 done
Networks that understand pictures
An ordinary layer connects every pixel to every neuron. For images that’s wasteful. A 1 000 × 1 000 photo would need a million weights per neuron. And it ignores two facts about pictures: nearby pixels belong together, and a cat is a cat wherever it sits in the frame.
A convolution uses both facts. It slides one small grid of weights, a kernel, across the whole image. At each position it takes the dot product of the kernel with the patch under it. The result is a new image, a feature map, that lights up wherever the patch looks like the kernel.