CNN Visualizer

Step through convolutional neural network layers � visualize feature maps, kernels, and pooling

Network Pipeline

?
?
?
?
?
?
?

Feature Maps � InputStep 1/8

Map
8�8

Image Input

Layer Info

TypeINPUT
Dimensions8�8�1

Dimensions Pipeline

Input8�8�1
Conv2D6�6�3
ReLU6�6�3
MaxPool3�3�3
Conv2D1�1�6
ReLU1�1�6
Fully Connected1�1�1
Output1�1�1

Summary

Input: 8�8 grayscale image (64 pixels)

Conv layers: 2 with 9 filters

Pooling: Reduces spatial dimensions by 2�2

Output: 3 class probabilities

How it works

Conv: Slides filters over input to produce feature maps.

ReLU: Applies max(0, x) element-wise for non-linearity.

Pool: Downsamples by taking max in each window.

FC: Connects all features to output neurons for classification.