deep learningcomputer visioninteractive · Aug 7, 2026 · 1 min read

Convolutions, Visualized

An interactive look at how a convolution kernel sweeps across an input tensor. Set your own input shape, kernel, stride, padding and dilation, and watch the output feature map take shape.

// interactive playground
step 0 / 0
Input height7
Input width7
Input channels3
Kernel size3
Stride1
Padding0
Dilation1
Output filters1
Animation speed
Input
Output
input cell zero-padding receptive field output cell
// kernel volume
// output volume
Output shape
out=⌊(in+2·paddilation·(k−1)1)/stride⌋+1
Height
Width
Channels
Layer weights
weights=C_in×k×k×C_out
Weights
Biases
Total params
// the same layer in pytorch
import torch
import torch.nn as nn

conv = nn.Conv2d(
    in_channels=3,
    out_channels=1,
    kernel_size=3,
    stride=1,
    padding=0,
    dilation=1,
    bias=True,
)

x = torch.randn(8, 3, 7, 7)
y = conv(x)