How to set the params of Deconvolution in Caffe prototxt?

[csyking]

Hi All,

If the size of feature map is 4*4, how to set the params of Deconvolution to make the feature map to 7*7?

The Deconvolution layer is as follows:

layer {

  name: "deconv5_2"

  type: "Deconvolution"

  bottom: "conv5_2"

  top: "deconv5_2"

  param {

    lr_mult: 1

    decay_mult: 0

  }

  param {

    lr_mult: 2

    decay_mult: 0

  }

  convolution_param {

    num_output: 512

    pad: 1 # is right?

    stride: 1 # is right?

    kernel_size: 6 # is right?

    weight_filler {

      type: "xavier"

    }

    bias_filler {

      type: "constant"

      value: 0

    }

  }

}

What is the mathematical formula of deconvolution？

Any help I will be grateful!

[Przemek D]

For convolution we have:
Y = (X + 2*p - k) / s + 1
Where s stands for stride, k for kernel size and p for padding, X is input size and Y for output. Deconvolution works exactly backwards, so all you want to do is reorganize the equation to free X:
X = s*(Y-1) - 2*p + k

[Ali]

Take a look at:

http://caffe.berkeleyvision.org/doxygen/classcaffe_1_1BilinearFiller.html