Skip to content

Latest commit

 

History

History
54 lines (36 loc) · 1.83 KB

GlobalMaxPoolingLayer.md

File metadata and controls

54 lines (36 loc) · 1.83 KB
Raw

CGlobalMaxPoolingLayer Class

This class implements a layer which performs max pooling on Height, Width, and Depth dimensions of the input, allowing for multiple largest elements to be found.

If you set the number of largest elements to 1, this layer will function exactly as the C3dMaxPoolingLayer with the filter size equal to the input size.

Settings

The number of maximum elements

void SetMaxCount(int enumSize);

Sets the number of largest elements that may be found. Note that these do not have to be equal to each other; the top GetMaxCount(); elements will be returned.

Trainable parameters

The layer has no trainable parameters.

Inputs

The single input accepts a blob of the following dimensions:

  • BatchLength * BatchWidth * ListSize - the number of images in the set
  • Height - the images' height
  • Width - the images' width
  • Depth - the images' depth
  • Channels - the number of channels the image format uses

Outputs

The layer may have one or two outputs. The first output contains the values found, in a blob of the dimensions:

  • BatchLength is equal to the input BatchLength
  • BatchWidth is equal to the input BatchWidth
  • ListSize is equal to the input ListSize
  • Height is equal to 1
  • Width is equal to GetMaxCount()
  • Depth is equal to 1
  • Channels is equal to the input Channels

The second output is optional. It has the same dimensions and contains the (integer) indices of the maximum values in the original blob.