Deconv2D

Warning

Documentation for the module is under development.

Description

Info

Parent class: DeconvND

Derived classes: -

This module performs the operation of 2-dimensional transposed convolution. For more detailed theoretical information about the transposed convolution operation, see DeconvND.

Initializing

def __init__(self, inmaps, outmaps, size, stride=1, pad=0, dilation=1, wscale=1.0, useBias=True, name=None,
                 initscheme=None, empty=False, groups=1):

Parameters

Parameter	Allowed types	Description	Default
inmaps	int	Number of maps in the input tensor	-
outmaps	int	Number of maps in the output tensor	-
size	int	Convolution kernel size (the kernel is always equilateral)	-
stride	Union[int, tuple]	Convolution stride	1
pad	Union[int, tuple]	Map padding	0
dilation	Union[int, tuple]	Convolution window dilation	1
wscale	float	Random layer weights variance	1.0
useBias	bool	Whether to use the bias vector	True
initscheme	Union[tuple, str]	Specifies the layer weights initialization scheme (see createTensorWithScheme)	None -> ("xavier_uniform", "in")
name	str	Layer name	None
empty	bool	Whether to initialize the matrix of weights and biases	False
groups	int	Number of groups the maps are split into for separate processing	1

Explanations

-

Examples

---

Basic deconvolution example

---

Necessary imports

import numpy as np
from PuzzleLib.Backend import gpuarray
from PuzzleLib.Modules import Deconv2D
from PuzzleLib.Variable import Variable

Info

gpuarray is required to properly place the tensor in the GPU.

batchsize, inmaps, h, w = 1, 2, 5, 5
outsize = 2

Synthetic tensor:

data = gpuarray.to_gpu(np.arange(batchsize * inmaps * h * w).reshape((batchsize, inmaps, h, w)).astype(np.float32))

deconv = Deconv2D(inmaps=inmaps, outmaps=outsize, size=2, useBias=False)
print(deconv(data))

---

Size parameter

---

deconv = Deconv2D(inmaps=inmaps,outmaps=outsize, size=3, useBias=False)
print(deconv(data))

---

Pad parameter

---

deconv = Deconv2D(inmaps=inmaps,outmaps=outsize, size=3, pad=1, useBias=False)
print(deconv(data))

---

Stride parameter

---

deconv = Deconv2D(inmaps=inmaps, outmaps=outsize, size=2, stride=2, useBias=False)
print(deconv(data))

deconv = Deconv2D(inmaps=inmaps, outmaps=outsize, size=2, stride=2, pad=3, useBias=False)
print(deconv(data))

deconv = Deconv2D(inmaps=inmaps, outmaps=outsize, size=2, stride=(2, 4), pad=3, useBias=False)
print(deconv(data))

---

Параметр dilation

---

deconv = Deconv2D(inmaps=inmaps, outmaps=outsize, size=2, stride=1, pad=0, dilation=2, useBias=False)
print(deconv(data))

deconv = Deconv2D(inmaps=inmaps, outmaps=outsize, size=2, stride=1, pad=0, dilation=(3, 1), useBias=False)
print(deconv(data))

---

Groups parameter

---

deconv = Deconv2D(inmaps=inmaps, outmaps=outsize, size=2, groups=2, useBias=False)
print(deconv(data))