LookingForSeagrassSemanticSegmentation/nnUtils.py at master · EnviewFulda/LookingForSeagrassSemanticSegmentation · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
import tensorflow as tf

STRIDE = 2

def weight_variable(shape, name):
      return tf.get_variable(name, shape=shape,
          initializer=tf.contrib.layers.xavier_initializer())#tf.initializers.orthogonal())

def bias_variable(shape, name):
    initial = tf.constant(0.1, shape=shape, name=name)
    return tf.Variable(initial)

def conv(input, filter, name, pad="SAME", dilation=0, dropR=0.3):
    f = weight_variable(filter, name+"f1")

    if not dilation > 0:
        conv = tf.nn.conv2d(input, f, strides=[1,1,1,1], padding=pad, name=name)
    else:
        conv = tf.nn.atrous_conv2d(input, f, dilation, padding=pad)

    #conv_bias = tf.nn.bias_add(conv, bias_variable([filter[3]], name=name+"b1"))
    batch_norm = tf.contrib.layers.batch_norm(conv)

    relu = tf.nn.relu(batch_norm)

    #if leakyR:
    #relu = tf.nn.leaky_relu(batch_norm,alpha=0.2,name=None)

    #drop = tf.nn.dropout(relu, dropR)

    return relu

def pool(input, window, stride, poolIndices=False, name="POOL"):


    if poolIndices:
        pool = tf.nn.max_pool_with_argmax(
                input,
                ksize=[1, window, window,1],
                strides=[1, stride, stride, 1],
                padding="SAME",
                name=name
            )
    else:
        pool = tf.nn.avg_pool(
                input,
                ksize=[1, window, window,1],
                strides=[1, stride, stride, 1],
                padding="SAME",
                name=name
            )

    return pool

def deconv_filter(shape, name):

    # filter = tf.zeros((
    #     shape[0], # height
    #     shape[1], # width
    #     shape[2], # out channels
    #     shape[3] # in channels
    #  ), name=name)

    #return tf.Variable(filter)
    return tf.get_variable(name, shape=shape,
           initializer=tf.contrib.layers.xavier_initializer())

def deconv(layer, outputShape, filterShape, name, stride=2):
    filter = deconv_filter(filterShape, "deconvF"+layer.name[:len(layer.name)-2])

    deconv = tf.nn.conv2d_transpose(
            layer,
            filter,
            outputShape,
            strides=[1,stride,stride,1],
            padding="SAME",
            name=name
        )

    #deconv = tf.image.resize_bilinear(
     #             layer, outputShape, align_corners=True)
    return deconv