File size: 5,279 Bytes
1ab1a09
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#    http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import paddle
import paddle.nn as nn
import paddle.nn.functional as F

from paddleseg.cvlibs import manager
from paddleseg.models import layers
from paddleseg.utils import utils


@manager.MODELS.add_component
class DMNet(nn.Layer):
    """
    The DMNet implementation based on PaddlePaddle.

    The original article refers to
     Junjun He, Zhongying Deng, Yu Qiao. "Dynamic Multi-scale Filters for Semantic Segmentation"

    Args:
        num_classes (int): The unique number of target classes.
        backbone (paddle.nn.Layer): Backbone network, currently support Resnet50_vd/Resnet101_vd.
        mid_channels (int): The middle channels of convolution layer. Default: 512.
        filter_sizes (list, tuple): The filter size of generated convolution kernel used in Dynamic Convolutional Module. Default: [1, 3, 5, 7].
        fusion (bool): Add one conv to fuse DCM output feature. Default: False.
        pretrained (str, optional): The path or url of pretrained model. Default: None.
    """

    def __init__(self,
                 num_classes,
                 backbone,
                 mid_channels=512,
                 filter_sizes=[1, 3, 5, 7],
                 fusion=False,
                 pretrained=None):
        super().__init__()
        self.backbone = backbone
        self.dcm_modules = nn.LayerList()
        for filter_size in filter_sizes:
            self.dcm_modules.append(
                DCM(filter_size, fusion, self.backbone.feat_channels[-1],
                    mid_channels), )
        self.bottleneck = layers.ConvBNReLU(
            self.backbone.feat_channels[-1] + len(filter_sizes) * mid_channels,
            mid_channels,
            3,
            padding=1, )
        self.cls = nn.Conv2D(mid_channels, num_classes, 1)

        self.fcn_head = nn.Sequential(
            layers.ConvBNReLU(
                self.backbone.feat_channels[2], mid_channels, 3, padding=1),
            nn.Conv2D(mid_channels, num_classes, 1), )

        self.pretrained = pretrained
        self.init_weight()

    def init_weight(self):
        if self.pretrained is not None:
            utils.load_entire_model(self, self.pretrained)

    def forward(self, x):
        feats = self.backbone(x)
        x = feats[-1]
        dcm_outs = [x]
        for dcm_module in self.dcm_modules:
            dcm_outs.append(dcm_module(x))
        dcm_outs = paddle.concat(dcm_outs, axis=1)
        x = self.bottleneck(dcm_outs)
        x = self.cls(x)
        x = F.interpolate(
            x, scale_factor=8, mode='bilinear', align_corners=True)
        output = [x]
        if self.training:
            fcn_out = self.fcn_head(feats[2])
            fcn_out = F.interpolate(
                fcn_out, scale_factor=8, mode='bilinear', align_corners=True)
            output.append(fcn_out)
            return output
        return output


class DCM(nn.Layer):
    """
    Dynamic Convolutional Module used in DMNet.

    Args:
        filter_size (int): The filter size of generated convolution kernel used in Dynamic Convolutional Module.
        fusion (bool): Add one conv to fuse DCM output feature.
        in_channels (int): Input channels.
        channels (int): Channels after modules, before conv_seg.
    """

    def __init__(self, filter_size, fusion, in_channels, channels):
        super().__init__()
        self.filter_size = filter_size
        self.fusion = fusion
        self.channels = channels

        pad = (self.filter_size - 1) // 2
        if (self.filter_size - 1) % 2 == 0:
            self.pad = (pad, pad, pad, pad)
        else:
            self.pad = (pad + 1, pad, pad + 1, pad)

        self.avg_pool = nn.AdaptiveAvgPool2D(filter_size)
        self.filter_gen_conv = nn.Conv2D(in_channels, channels, 1)
        self.input_redu_conv = layers.ConvBNReLU(in_channels, channels, 1)

        self.norm = layers.SyncBatchNorm(channels)
        self.act = nn.ReLU()

        if self.fusion:
            self.fusion_conv = layers.ConvBNReLU(channels, channels, 1)

    def forward(self, x):
        generated_filter = self.filter_gen_conv(self.avg_pool(x))
        x = self.input_redu_conv(x)
        b, c, h, w = x.shape
        x = x.reshape([1, b * c, h, w])
        generated_filter = generated_filter.reshape(
            [b * c, 1, self.filter_size, self.filter_size])

        x = F.pad(x, self.pad, mode='constant', value=0)
        output = F.conv2d(x, weight=generated_filter, groups=b * c)
        output = output.reshape([b, self.channels, h, w])
        output = self.norm(output)
        output = self.act(output)
        if self.fusion:
            output = self.fusion_conv(output)
        return output