AI Foundationsintermediate13 min · 9 steps
Convolution & Pooling
The two operations inside every CNN: a small learned kernel slid across the image (2D and depthwise convolution), and pooling that shrinks feature maps by keeping the max or the average.
weights150.5M
heavy workdataresponse
metrics are illustrativeProblemStep 1 of 9
A dense layer on an image is huge and naive
A 224×224 colour image has 224 × 224 × 3 = 150,528 numbers. Connect it to a dense layer of 1,000 units and that single layer needs about 150 million weights.
Worse, it learns nothing reusable: a cat in the top-left corner and the same cat in the centre hit completely different weights, and the layer ignores the fact that neighbouring pixels belong together.
Lesson 19 of 34 in AI Foundations