Third Party

This module contains third party code that I used from outside in support of this.

This file contains third party support code that I have used from elsewhere. I have cited them appropriately where ever needed

lenet.third_party.put_kernels_on_grid(kernel, pad=1, name='visualizer')

Visualize convolutional filters as an image (mostly for the 1st layer). Arranges filters into a grid, with some paddings between adjacent filters.

Parameters:

• kernel – tensor of shape [Y, X, NumChannels, NumKernels] (HWCN)
• pad – number of black pixels around each filter (between them)
• name – name for tensorflow scope

Returns:

Tensor of shape [1, (Y+2*pad)*grid_Y, (X+2*pad)*grid_X, NumChannels].

Notes

This is not my method. This was written by kukurza and was hosted at: https://gist.github.com/kukuruza/03731dc494603ceab0c5