File size: 7,456 Bytes
f549064
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
# Copyright (c) OpenMMLab. All rights reserved.
from typing import List, Tuple

import torch
import torch.nn.functional as F
from mmcv.cnn import ConvModule
from mmengine.model import BaseModule

from mmdet.registry import MODELS
from mmdet.utils import ConfigType, OptConfigType, OptMultiConfig


class SSHContextModule(BaseModule):
    """This is an implementation of `SSH context module` described in `SSH:
    Single Stage Headless Face Detector.

    <https://arxiv.org/pdf/1708.03979.pdf>`_.

    Args:
        in_channels (int): Number of input channels used at each scale.
        out_channels (int): Number of output channels used at each scale.
        conv_cfg (:obj:`ConfigDict` or dict, optional): Config dict for
            convolution layer. Defaults to None.
        norm_cfg (:obj:`ConfigDict` or dict): Config dict for normalization
            layer. Defaults to dict(type='BN').
        init_cfg (:obj:`ConfigDict` or list[:obj:`ConfigDict`] or dict or
            list[dict], optional): Initialization config dict.
            Defaults to None.
    """

    def __init__(self,
                 in_channels: int,
                 out_channels: int,
                 conv_cfg: OptConfigType = None,
                 norm_cfg: ConfigType = dict(type='BN'),
                 init_cfg: OptMultiConfig = None):
        super().__init__(init_cfg=init_cfg)
        assert out_channels % 4 == 0

        self.in_channels = in_channels
        self.out_channels = out_channels

        self.conv5x5_1 = ConvModule(
            self.in_channels,
            self.out_channels // 4,
            3,
            stride=1,
            padding=1,
            conv_cfg=conv_cfg,
            norm_cfg=norm_cfg,
        )

        self.conv5x5_2 = ConvModule(
            self.out_channels // 4,
            self.out_channels // 4,
            3,
            stride=1,
            padding=1,
            conv_cfg=conv_cfg,
            norm_cfg=norm_cfg,
            act_cfg=None)

        self.conv7x7_2 = ConvModule(
            self.out_channels // 4,
            self.out_channels // 4,
            3,
            stride=1,
            padding=1,
            conv_cfg=conv_cfg,
            norm_cfg=norm_cfg,
        )

        self.conv7x7_3 = ConvModule(
            self.out_channels // 4,
            self.out_channels // 4,
            3,
            stride=1,
            padding=1,
            conv_cfg=conv_cfg,
            norm_cfg=norm_cfg,
            act_cfg=None,
        )

    def forward(self, x: torch.Tensor) -> tuple:
        conv5x5_1 = self.conv5x5_1(x)
        conv5x5 = self.conv5x5_2(conv5x5_1)
        conv7x7_2 = self.conv7x7_2(conv5x5_1)
        conv7x7 = self.conv7x7_3(conv7x7_2)

        return (conv5x5, conv7x7)


class SSHDetModule(BaseModule):
    """This is an implementation of `SSH detection module` described in `SSH:
    Single Stage Headless Face Detector.

    <https://arxiv.org/pdf/1708.03979.pdf>`_.

    Args:
        in_channels (int): Number of input channels used at each scale.
        out_channels (int): Number of output channels used at each scale.
        conv_cfg (:obj:`ConfigDict` or dict, optional): Config dict for
            convolution layer. Defaults to None.
        norm_cfg (:obj:`ConfigDict` or dict): Config dict for normalization
            layer. Defaults to dict(type='BN').
        init_cfg (:obj:`ConfigDict` or list[:obj:`ConfigDict`] or dict or
            list[dict], optional): Initialization config dict.
            Defaults to None.
    """

    def __init__(self,
                 in_channels: int,
                 out_channels: int,
                 conv_cfg: OptConfigType = None,
                 norm_cfg: ConfigType = dict(type='BN'),
                 init_cfg: OptMultiConfig = None):
        super().__init__(init_cfg=init_cfg)
        assert out_channels % 4 == 0

        self.in_channels = in_channels
        self.out_channels = out_channels

        self.conv3x3 = ConvModule(
            self.in_channels,
            self.out_channels // 2,
            3,
            stride=1,
            padding=1,
            conv_cfg=conv_cfg,
            norm_cfg=norm_cfg,
            act_cfg=None)

        self.context_module = SSHContextModule(
            in_channels=self.in_channels,
            out_channels=self.out_channels,
            conv_cfg=conv_cfg,
            norm_cfg=norm_cfg)

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        conv3x3 = self.conv3x3(x)
        conv5x5, conv7x7 = self.context_module(x)
        out = torch.cat([conv3x3, conv5x5, conv7x7], dim=1)
        out = F.relu(out)

        return out


@MODELS.register_module()
class SSH(BaseModule):
    """`SSH Neck` used in `SSH: Single Stage Headless Face Detector.

    <https://arxiv.org/pdf/1708.03979.pdf>`_.

    Args:
        num_scales (int): The number of scales / stages.
        in_channels (list[int]): The number of input channels per scale.
        out_channels (list[int]): The number of output channels  per scale.
        conv_cfg (:obj:`ConfigDict` or dict, optional): Config dict for
            convolution layer. Defaults to None.
        norm_cfg (:obj:`ConfigDict` or dict): Config dict for normalization
            layer. Defaults to dict(type='BN').
        init_cfg (:obj:`ConfigDict` or list[:obj:`ConfigDict`] or dict or
            list[dict], optional): Initialization config dict.

    Example:
        >>> import torch
        >>> in_channels = [8, 16, 32, 64]
        >>> out_channels = [16, 32, 64, 128]
        >>> scales = [340, 170, 84, 43]
        >>> inputs = [torch.rand(1, c, s, s)
        ...           for c, s in zip(in_channels, scales)]
        >>> self = SSH(num_scales=4, in_channels=in_channels,
        ...           out_channels=out_channels)
        >>> outputs = self.forward(inputs)
        >>> for i in range(len(outputs)):
        ...     print(f'outputs[{i}].shape = {outputs[i].shape}')
        outputs[0].shape = torch.Size([1, 16, 340, 340])
        outputs[1].shape = torch.Size([1, 32, 170, 170])
        outputs[2].shape = torch.Size([1, 64, 84, 84])
        outputs[3].shape = torch.Size([1, 128, 43, 43])
    """

    def __init__(self,
                 num_scales: int,
                 in_channels: List[int],
                 out_channels: List[int],
                 conv_cfg: OptConfigType = None,
                 norm_cfg: ConfigType = dict(type='BN'),
                 init_cfg: OptMultiConfig = dict(
                     type='Xavier', layer='Conv2d', distribution='uniform')):
        super().__init__(init_cfg=init_cfg)
        assert (num_scales == len(in_channels) == len(out_channels))
        self.num_scales = num_scales
        self.in_channels = in_channels
        self.out_channels = out_channels

        for idx in range(self.num_scales):
            in_c, out_c = self.in_channels[idx], self.out_channels[idx]
            self.add_module(
                f'ssh_module{idx}',
                SSHDetModule(
                    in_channels=in_c,
                    out_channels=out_c,
                    conv_cfg=conv_cfg,
                    norm_cfg=norm_cfg))

    def forward(self, inputs: Tuple[torch.Tensor]) -> tuple:
        assert len(inputs) == self.num_scales

        outs = []
        for idx, x in enumerate(inputs):
            ssh_module = getattr(self, f'ssh_module{idx}')
            out = ssh_module(x)
            outs.append(out)

        return tuple(outs)