Watch a real 3×3 filter slide, multiply, and sum its way across an actual image.
📋 What Am I Looking At? (Read This First)
The glowing 3×3 box is the Kernel (also called a Filter) — a small grid of learned numbers, NOT part of the image.
Wherever the Kernel sits, that 3×3 patch of the image underneath it is called its Receptive Field — the only pixels it can currently "see."
Convolution = multiply each pixel by the matching Kernel number, then add all 9 results — that single sum is called the Dot Product.
Every dot product becomes ONE pixel in the Feature Map on the right — the network's summary of "how strongly this pattern appeared here."
Press ▶ Convolve to watch it slide the whole way across, one Receptive Field at a time.
Receptive Field: row 0, col 0
Dot Product = —
Pick a Kernel Type below, then press ▶ Convolve to begin.
Pixel-Math Overlay Off
On
🔍 SCANNERPress ▶ Convolve — I'll slide this Kernel across every possible position and build the Feature Map live.
🧑🏫 PROF. TORCHA Kernel doesn't know what an "edge" is in advance — its NUMBERS are just arranged so edges happen to produce a big dot product and flat regions produce zero.