CNN Receptive Field Growth Across Layers

Network Depth & Filters

Convolutional layers L 4

Kernel size k×k 3×3

Stride s 1

Dilation d 1

Live Readouts

Receptive field

7×7

Output map size

26×26

Cumulative stride (jump)

1 px / output step

Legend

Selected output neuron (top layer) Cells inside the receptive field Cells outside the receptive field

How it works

Each stacked plane is one convolutional layer's feature map, drawn on a 32×32 input. A conv layer with kernel k, stride s and dilation d shrinks the spatial size:

k_eff = d·(k − 1) + 1
n_out = floor((n_in − k_eff) / s) + 1

The receptive field is the region of the original input that can influence one output unit. It grows layer by layer via the standard recurrence (RF₀ = 1, jump₀ = 1):

RF_l   = RF_(l-1) + (k_eff − 1) · jump_(l-1)
jump_l = jump_(l-1) · s

This simulation picks one output neuron at the very top layer (yellow) and walks the receptive-field recurrence backward through every layer, highlighting in red exactly which cells of each feature map — and ultimately which patch of the raw input — feed into that single neuron's value.

Layers L — how many conv layers are stacked; more layers compound the receptive field.
Kernel size k — a larger filter directly widens the field at every layer.
Stride s — shrinks feature maps fastest and multiplies the "jump", so later layers cover disproportionately more input per step.
Dilation d — spreads a kernel's taps apart (à la WaveNet / DeepLab atrous convolution) without adding parameters, growing k_eff and the receptive field for free.

Real-world relevance: this exact recurrence is why deep CNNs (ResNet, VGG) need only 3×3 kernels to eventually "see" the whole image, and why dilated convolutions are used in segmentation networks to grow context cheaply.

⚠ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.