Pooling Output Size Calculator - AI & Machine Learning

Calculator Form

Use the inputs below to estimate pooling layer output size for 1D, 2D, or 3D data.

Responsive form: 3 columns large, 2 medium, 1 mobile

Pooling Type

Dimension

Batch Size

Channels / Feature Maps

Use Ceil Mode

Count Padding in Average Divisor

Depth Axis

Input Depth

Kernel Depth

Stride Depth

Front Padding

Back Padding

Depth Dilation

Height Axis

Input Height

Kernel Height

Stride Height

Top Padding

Bottom Padding

Height Dilation

Width / Length Axis

Input Width / Length

Kernel Width / Length

Stride Width / Length

Left / Start Padding

Right / End Padding

Width / Length Dilation

Reset

Example Data Table

These examples show common pooling setups for sequence, image, and volume data.

Use Case	Dimension	Input Shape	Kernel	Stride	Padding	Dilation	Mode	Output Shape
Text or signal pooling	1D	128	4	2	1 + 1	1	Floor	64
Image feature map reduction	2D	32 × 32	2 × 2	2 × 2	0 + 0 / 0 + 0	1 × 1	Floor	16 × 16
Volume segmentation model	3D	16 × 64 × 64	2 × 2 × 2	2 × 2 × 2	0 + 0 on all axes	1 × 1 × 1	Floor	8 × 32 × 32
Asymmetric pooled image	2D	31 × 45	3 × 3	2 × 2	1 + 1 / 0 + 2	1 × 1	Ceil	16 × 23

Formula Used

For each active axis, the calculator first computes the effective kernel. It then evaluates the raw output length and applies either floor mode or ceil mode.

Effective Kernel

effective_kernel = dilation × (kernel − 1) + 1

Floor Mode Output

output = floor(((input + pad_before + pad_after − effective_kernel) / stride) + 1)

Ceil Mode Output

output = ceil(((input + pad_before + pad_after − effective_kernel) / stride) + 1)

Volume and Reduction

input_volume = product of all active input axes output_volume = product of all active output axes reduction_factor = input_volume / output_volume

The pooling type does not change the output size formula. It changes how values inside each pooling window are aggregated.

How to Use This Calculator

Choose the pooling type and select 1D, 2D, or 3D.
Enter batch size and channel count if you want total activation counts.
Fill in the input size, kernel, stride, padding, and dilation for each active axis.
Enable ceil mode when your framework uses rounded-up output sizing.
Click Calculate Output Size to show the result above the form.
Review the summary, per-axis table, and graph for dimensional reduction.
Use the CSV and PDF buttons to export the calculated report.

Frequently Asked Questions

1) What does pooling output size mean?

It is the size of the feature map after a pooling layer slides across the input using the selected kernel, stride, padding, and dilation values.

2) Does max pooling and average pooling change output dimensions?

No. The output shape formula is the same for both. Only the way each pooling window combines values changes between the two methods.

3) Why is dilation included here?

Dilation increases the effective kernel span by inserting gaps between sampled positions. It changes the output size because the pooling window covers a wider area.

4) What is effective kernel size?

Effective kernel size is the real span seen by the layer after dilation is applied. It equals dilation × (kernel − 1) + 1.

5) When should I use ceil mode?

Use ceil mode when your framework or model definition rounds up partial window coverage. It can produce larger output dimensions than standard floor mode.

6) Why do I see zero as an output size?

That happens when the kernel span is too large for the padded input or when the configuration leaves no valid pooling window position.

7) Can this calculator handle 3D medical or video tensors?

Yes. Select 3D and enter depth, height, and width values. The calculator then computes output sizes across all three spatial axes.

8) What does reduction factor help me understand?

It shows how strongly the layer downsamples data. A higher factor means stronger spatial reduction and fewer output activations to process later.