-
Notifications
You must be signed in to change notification settings - Fork 9
/
Copy pathmgn_ptl.py
204 lines (182 loc) · 8.54 KB
/
mgn_ptl.py
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
import copy
import torch
import torch.nn as nn
from opt import opt
from torchvision.models.resnet import resnet50, resnet101, Bottleneck
from bconv_cell import BConvCell
import torch.nn.init as init
class PTL(nn.Module):
""" Init the Progressive Transfer Learning Network.
'Progressive transfer learning for person re-identification' by Yu et al.
"""
def __init__(self):
super(PTL, self).__init__()
self.bconv1 = BConvCell(3, 16, kernel_size=3, stride=2, padding=1)
self.bconv2 = BConvCell(16, 32, kernel_size=3, stride=2, padding=1)
self.bconv3 = BConvCell(64, 64, kernel_size=3, padding=1)
self.bconv4 = BConvCell(128, 256, kernel_size=3, stride=2, padding=1)
self.bconv5 = BConvCell(256, 512, kernel_size=3, stride=2, padding=1)
self.bconv6 = BConvCell(512, 128, kernel_size=3, stride=2, padding=1)
self.bconv7 = BConvCell(512, 128, kernel_size=3, padding=1)
self.bconv8 = BConvCell(512, 128, kernel_size=3, padding=1)
self.fuse1 = nn.Sequential(
nn.Conv2d(64 + 32, 64, 1, bias=True),
nn.BatchNorm2d(64),
nn.ReLU(True)
)
self.fuse2 = nn.Sequential(
nn.Conv2d(256 + 64, 128, 1, bias=True),
nn.BatchNorm2d(128),
nn.ReLU(True)
)
self.fuse3 = nn.Sequential(
nn.Conv2d(512 + 256, 256, 1, bias=True),
nn.BatchNorm2d(256),
nn.ReLU(True)
)
self.fuse4 = nn.Sequential(
nn.Conv2d(1024 + 512, 512, 1, bias=True),
nn.BatchNorm2d(512),
nn.ReLU(True)
)
self.maxpool_zg_p1 = nn.MaxPool2d(kernel_size=(12, 4))
self.maxpool_zg_p2 = nn.MaxPool2d(kernel_size=(24, 8))
self.maxpool_zg_p3 = nn.MaxPool2d(kernel_size=(24, 8))
self.maxpool_zp2 = nn.MaxPool2d(kernel_size=(12, 8))
self.maxpool_zp3 = nn.MaxPool2d(kernel_size=(8, 8))
def resetalllatentstates(self):
self.bconv1.resetlatentstate()
self.bconv2.resetlatentstate()
self.bconv3.resetlatentstate()
self.bconv4.resetlatentstate()
self.bconv5.resetlatentstate()
self.bconv6.resetlatentstate()
self.bconv7.resetlatentstate()
self.bconv8.resetlatentstate()
def init_param(self):
print('\nInit PTL Net Params\n')
for m in self.modules():
if isinstance(m, nn.Conv2d):
init.kaiming_normal_(m.weight, mode='fan_in')
class MGN_PTL(nn.Module):
""" Init the Multiple Granularities Network.
Implement of paper: 'Learning Discriminative Features with Multiple Granularities for Person Re-Identification' proposed by Wang et al.
Refer url: https://github.com/GNAYUOHZ/ReID-MGN
"""
def __init__(self):
super(MGN_PTL, self).__init__()
num_classes = opt.classn
feats = 256
self.val = False
self.ptl = PTL()
self.ptl.init_param()
if opt.backbone == 'resnet50':
self.backbone = resnet50(pretrained=True)
elif opt.backbone == 'resnet101':
self.backbone = resnet101(pretrained=True)
res_conv4 = nn.Sequential(*self.backbone.layer3[1:])
res_g_conv5 = self.backbone.layer4
res_p_conv5 = nn.Sequential(
Bottleneck(1024, 512, downsample=nn.Sequential(nn.Conv2d(1024, 2048, 1, bias=False), nn.BatchNorm2d(2048))),
Bottleneck(2048, 512),
Bottleneck(2048, 512))
res_p_conv5.load_state_dict(self.backbone.layer4.state_dict())
self.p1 = nn.Sequential(copy.deepcopy(res_conv4), copy.deepcopy(res_g_conv5))
self.p2 = nn.Sequential(copy.deepcopy(res_conv4), copy.deepcopy(res_p_conv5))
self.p3 = nn.Sequential(copy.deepcopy(res_conv4), copy.deepcopy(res_p_conv5))
self.maxpool_zg_p1 = nn.MaxPool2d(kernel_size=(12, 4))
self.maxpool_zg_p2 = nn.MaxPool2d(kernel_size=(24, 8))
self.maxpool_zg_p3 = nn.MaxPool2d(kernel_size=(24, 8))
self.maxpool_zp2 = nn.MaxPool2d(kernel_size=(12, 8))
self.maxpool_zp3 = nn.MaxPool2d(kernel_size=(8, 8))
self.reduction = nn.Sequential(nn.Conv2d(2048+128, feats, 1, bias=True), nn.BatchNorm2d(feats), nn.ReLU())
self._init_reduction(self.reduction)
self.fc_id_2048_0 = nn.Linear(feats, num_classes)
self.fc_id_2048_1 = nn.Linear(feats, num_classes)
self.fc_id_2048_2 = nn.Linear(feats, num_classes)
self.fc_id_256_1_0 = nn.Linear(feats, num_classes)
self.fc_id_256_1_1 = nn.Linear(feats, num_classes)
self.fc_id_256_2_0 = nn.Linear(feats, num_classes)
self.fc_id_256_2_1 = nn.Linear(feats, num_classes)
self.fc_id_256_2_2 = nn.Linear(feats, num_classes)
self._init_fc(self.fc_id_2048_0)
self._init_fc(self.fc_id_2048_1)
self._init_fc(self.fc_id_2048_2)
self._init_fc(self.fc_id_256_1_0)
self._init_fc(self.fc_id_256_1_1)
self._init_fc(self.fc_id_256_2_0)
self._init_fc(self.fc_id_256_2_1)
self._init_fc(self.fc_id_256_2_2)
@staticmethod
def _init_reduction(reduction):
# conv
nn.init.kaiming_normal_(reduction[0].weight, mode='fan_in')
# bn
nn.init.normal_(reduction[1].weight, mean=1., std=0.02)
nn.init.constant_(reduction[1].bias, 0.)
@staticmethod
def _init_fc(fc):
nn.init.kaiming_normal_(fc.weight, mode='fan_in')
nn.init.constant_(fc.bias, 0.)
def resetalllatentstates(self):
self.ptl.resetalllatentstates()
def forward(self, input):
if self.val:
self.resetalllatentstates()
x1 = self.backbone.maxpool(self.backbone.relu(self.backbone.bn1(self.backbone.conv1(input))))
bconv_1 = self.ptl.bconv1(input)
bconv_2 = self.ptl.bconv2(bconv_1)
z0 = self.ptl.fuse1(torch.cat([bconv_2, x1], dim=1))
bconv_3 = self.ptl.bconv3(z0)
x2 = self.backbone.layer1(x1)
z1 = self.ptl.fuse2(torch.cat([x2, bconv_3], dim=1))
bconv_4 = self.ptl.bconv4(z1)
x3 = self.backbone.layer2(x2)
z2 = self.ptl.fuse3(torch.cat([x3, bconv_4], dim=1))
bconv_5 = self.ptl.bconv5(z2)
x4 = self.backbone.layer3[0](x3)
p1 = self.p1(x4)
p2 = self.p2(x4)
p3 = self.p3(x4)
z3 = self.ptl.fuse4(torch.cat([x4, bconv_5], dim=1))
bconv_6 = self.ptl.bconv6(z3)
bconv_7 = self.ptl.bconv7(z3)
bconv_8 = self.ptl.bconv8(z3)
zg_p1 = self.maxpool_zg_p1(p1)
zg_p2 = self.maxpool_zg_p2(p2)
zg_p3 = self.maxpool_zg_p3(p3)
zg_bconv_6 = self.ptl.maxpool_zg_p1(bconv_6)
zg_bconv_7 = self.ptl.maxpool_zg_p2(bconv_7)
zg_bconv_8 = self.ptl.maxpool_zg_p3(bconv_8)
zp2_bconv_7 = self.ptl.maxpool_zp2(bconv_7)
z0_p2_bconv_7 = zp2_bconv_7[:, :, 0:1, :]
z1_p2_bconv_7 = zp2_bconv_7[:, :, 1:2, :]
zp2 = self.maxpool_zp2(p2)
z0_p2 = zp2[:, :, 0:1, :]
z1_p2 = zp2[:, :, 1:2, :]
zp2_bconv_8 = self.ptl.maxpool_zp3(bconv_8)
z0_p3_bconv_8 = zp2_bconv_8[:, :, 0:1, :]
z1_p3_bconv_8 = zp2_bconv_8[:, :, 1:2, :]
z2_p3_bconv_8 = zp2_bconv_8[:, :, 2:3, :]
zp3 = self.maxpool_zp3(p3)
z0_p3 = zp3[:, :, 0:1, :]
z1_p3 = zp3[:, :, 1:2, :]
z2_p3 = zp3[:, :, 2:3, :]
fg_p1 = self.reduction(torch.cat([zg_p1, zg_bconv_6], dim=1)).squeeze(dim=3).squeeze(dim=2)
fg_p2 = self.reduction(torch.cat([zg_p2, zg_bconv_7], dim=1)).squeeze(dim=3).squeeze(dim=2)
fg_p3 = self.reduction(torch.cat([zg_p3, zg_bconv_8], dim=1)).squeeze(dim=3).squeeze(dim=2)
f0_p2 = self.reduction(torch.cat([z0_p2, z0_p2_bconv_7], dim=1)).squeeze(dim=3).squeeze(dim=2)
f1_p2 = self.reduction(torch.cat([z1_p2, z1_p2_bconv_7], dim=1)).squeeze(dim=3).squeeze(dim=2)
f0_p3 = self.reduction(torch.cat([z0_p3, z0_p3_bconv_8], dim=1)).squeeze(dim=3).squeeze(dim=2)
f1_p3 = self.reduction(torch.cat([z1_p3, z1_p3_bconv_8], dim=1)).squeeze(dim=3).squeeze(dim=2)
f2_p3 = self.reduction(torch.cat([z2_p3, z2_p3_bconv_8], dim=1)).squeeze(dim=3).squeeze(dim=2)
l_p1 = self.fc_id_2048_0(fg_p1)
l_p2 = self.fc_id_2048_1(fg_p2)
l_p3 = self.fc_id_2048_2(fg_p3)
l0_p2 = self.fc_id_256_1_0(f0_p2)
l1_p2 = self.fc_id_256_1_1(f1_p2)
l0_p3 = self.fc_id_256_2_0(f0_p3)
l1_p3 = self.fc_id_256_2_1(f1_p3)
l2_p3 = self.fc_id_256_2_2(f2_p3)
predict = torch.cat([fg_p1, fg_p2, fg_p3, f0_p2, f1_p2, f0_p3, f1_p3, f2_p3], dim=1)
return predict, fg_p1, fg_p2, fg_p3, l_p1, l_p2, l_p3, l0_p2, l1_p2, l0_p3, l1_p3, l2_p3