Computing & Software

Convolution Output Shape & Parameter Calculator

Calculate standard 2D convolution dimensions, parameter count and nominal multiply-accumulate work.

How the calculation works

Output feature-map height = floor((height+2*padding-dilation*(kernel-1)-1)/stride+1) pixels. Output feature-map width = floor((width+2*padding-dilation*(kernel-1)-1)/stride+1) pixels. Weights plus per-output-channel bias = (kernel**2*in_channels+1)*out_channels parameters. Nominal multiply-accumulates per image = floor((height+2*padding-dilation*(kernel-1)-1)/stride+1)*floor((width+2*padding-dilation*(kernel-1)-1)/stride+1)*out_channels*in_channels*kernel**2 MACs.

Worked example

Example inputs — Input feature-map height: 32; Input feature-map width: 32; Input channels: 3; Output channels / filters: 16; Square kernel size: 3; Same stride in both dimensions: 1; Padding on each side: 1; Kernel dilation: 1. Expected results — Output feature-map height: 32 pixels; Output feature-map width: 32 pixels; Weights plus per-output-channel bias: 448 parameters; Nominal multiply-accumulates per image: 442368 MACs.

Assumptions and limits

Analytical software/hardware scenario, not a benchmark or a live provider quote. Use the actual architecture, workload and billing definitions; overheads not named in the model are excluded. Square kernel, equal stride/padding, one group, one image and bias enabled. No pooling, transposed convolution, grouped/depthwise connectivity or implementation-specific FLOP convention is assumed.

Reference and further reading

Read the underlying subject reference. Calculith is an independent implementation; the reference does not endorse or certify this tool.

Related calculators

Browse all 1000 calculators →

Example values are editable. Results are estimates where assumptions apply. Calculations run in your browser. Updated 2026-11-16.