Convolution Visualizer

Dynamic heatmap visualization of feature extraction in deep learning.

Filter Presets

Filter Weights (3x3)

Step Calculation

Click "Next Step" to start...

Input Image 7x7

Values: 0 (Black) to 255 (White)

Filter

Feature Map 5x5

Heatmap based on result value

What is the color showing?

Input: Grayscale intensity. White (255) represents high activation.

Output: A dynamic heatmap. Green indicates high positive values, Red indicates negative values, and White/Gray is near zero.

Why use this?

The colors highlight how the filter "responds" to patterns. An edge detector creates a bright spot (high value) exactly where a sharp color change occurs in the input.