-
Notifications
You must be signed in to change notification settings - Fork 3.8k
/
Copy pathmodel.py
152 lines (127 loc) · 5.73 KB
/
model.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
import torch as t
import torch.nn as nn
import numpy as np
from torchvision.models import resnet18
from utils import get_loss
class ConvLayer(nn.Module):
def __init__(self, in_channels, out_channels, kernel_size, padding):
super(ConvLayer, self).__init__()
self.conv2d = nn.Sequential(
nn.Conv2d(in_channels,out_channels,kernel_size,stride=1,padding=padding,dilation=1),
nn.BatchNorm2d(out_channels),
nn.ReLU(inplace=True)
)
def forward(self, x):
return self.conv2d(x)
class DeConvLayer(nn.Module):
def __init__(self, in_channels, out_channels, kernel_size, stride):
super(DeConvLayer, self).__init__()
self.conv2d = nn.Sequential(
nn.ConvTranspose2d(in_channels,out_channels,kernel_size,stride=stride,padding=1,output_padding=0),
nn.BatchNorm2d(out_channels),
nn.ReLU(inplace=True)
)
def forward(self,x):
return self.conv2d(x)
class SPP(nn.Module):
def __init__(self):
super(SPP, self).__init__()
def forward(self,x):
x_1 = nn.functional.max_pool2d(x, 5, stride=1, padding=2)
x_2 = nn.functional.max_pool2d(x, 9, stride=1, padding=4)
x_3 = nn.functional.max_pool2d(x, 13, stride=1, padding=6)
return t.cat([x,x_1,x_2,x_3],dim=1)
class Model(nn.Module):
def __init__(self,num_classes,topk):
super(Model, self).__init__()
self.num_classes = num_classes
self.topk = topk
self.backbone = resnet18(pretrained=True)
self.backbone=nn.Sequential(*list(self.backbone.children())[:-2])
self.smooth = nn.Sequential(
SPP(),
ConvLayer(512*4,256,kernel_size=1,padding=0),
ConvLayer(256,512,kernel_size=3,padding=1)
)
self.deconv5 = DeConvLayer(512,256,kernel_size=4,stride=2)
self.deconv4 = DeConvLayer(256,256,kernel_size=4,stride=2)
self.deconv3 = DeConvLayer(256,256,kernel_size=4,stride=2)
self.cls_pred = nn.Sequential(
ConvLayer(256,64,kernel_size=3,padding=1),
nn.Conv2d(64,self.num_classes,kernel_size=1)
)
self.txty_pred = nn.Sequential(
ConvLayer(256,64,kernel_size=3,padding=1),
nn.Conv2d(64,2,kernel_size=1)
)
self.twth_pred = nn.Sequential(
ConvLayer(256,64,kernel_size=3,padding=1),
nn.Conv2d(64,2,kernel_size=1)
)
def decode(self,pred):
output = t.zeros_like(pred)
grid_y , grid_x = t.meshgrid([t.arange(128,device=pred.device),t.arange(128,device=pred.device)])
grid_cell = t.stack([grid_x,grid_y],dim=-1).float().view(1, 128*128, 2)
pred[:,:,:2] = (t.sigmoid(pred[:,:,:2])+grid_cell)*4
pred[:,:,2:] = (t.exp(pred[:,:,2:]))*4
#坐标转换：[cx,cy,w,h] -> [xmin,ymin,xmax,ymax]
output[:,:,0]=pred[:,:,0]-pred[:,:,2]/2
output[:,:,1]=pred[:,:,1]-pred[:,:,3]/2
output[:,:,2]=pred[:,:,0]+pred[:,:,2]/2
output[:,:,3]=pred[:,:,1]+pred[:,:,3]/2
return output
def gather_feat(self,feat,ind):
dim = feat.size(2)
ind = ind.unsqueeze(2).expand(ind.size(0),ind.size(1),dim)
return feat.gather(1,ind)
# 选取topk个满足要求的点
def get_topk(self,scores):
B,C,H,W = scores.size()
topk_scores,topk_inds = t.topk(scores.view(B,C,-1),self.topk)
topk_inds = topk_inds % (H*W)
topk_score,topk_ind = t.topk(topk_scores.view(B,-1),self.topk)
topk_inds = self.gather_feat(topk_inds.view(B,-1,1),topk_ind).view(B,self.topk)
topk_clses = t.floor_divide(topk_ind,self.topk).int()
return topk_score,topk_inds,topk_clses
def generate(self,x):
c5=self.backbone(x)
B=c5.size(0)
p5 = self.smooth(c5)
p4 = self.deconv5(p5)
p3 = self.deconv4(p4)
p2 = self.deconv3(p3)
cls_pred = self.cls_pred(p2)
txty_pred = self.txty_pred(p2)
twth_pred = self.twth_pred(p2)
cls_pred = t.sigmoid(cls_pred)
# 寻找8-近邻极大值点，其中keep为hmax极大值点的位置，cls_pred为对应的极大值点
hmax = nn.functional.max_pool2d(cls_pred,kernel_size=5, padding=2,stride=1)
keep = (hmax==cls_pred).float()
cls_pred *= keep
txtytwth_pred = t.cat([txty_pred,twth_pred],dim=1).permute(0,2,3,1).contiguous().view(B,-1,4)
scale = np.array([[[512,512,512,512]]])
scale_t = t.tensor(scale.copy(),device=txtytwth_pred.device).float()
bbox_pred = t.clamp((self.decode(txtytwth_pred)/scale_t)[0],0.,1.)
# 得到topk取值，topk_score：置信度，topk_ind：index，topk_clses：类别
topk_score,topk_ind,topk_clses = self.get_topk(cls_pred)
topk_bbox_pred = bbox_pred[topk_ind[0]]
return topk_bbox_pred.cpu().numpy(),topk_score[0].cpu().numpy(),topk_clses[0].cpu().numpy()
def forward(self,x,target):
c5=self.backbone(x)
B=c5.size(0)
p5 = self.smooth(c5)
p4 = self.deconv5(p5)
p3 = self.deconv4(p4)
p2 = self.deconv3(p3)
cls_pred = self.cls_pred(p2)
txty_pred = self.txty_pred(p2)
twth_pred = self.twth_pred(p2)
# 热力图：[B, H*W, num_classes]
cls_pred = cls_pred.permute(0, 2, 3, 1).contiguous().view(B, -1, self.num_classes)
# 中心点偏移：[B, H*W, 2]
txty_pred = txty_pred.permute(0, 2, 3, 1).contiguous().view(B, -1, 2)
# 物体尺度：[B, H*W, 2]
twth_pred = twth_pred.permute(0, 2, 3, 1).contiguous().view(B, -1, 2)
# 计算损失函数
total_loss = get_loss(pred_cls=cls_pred, pred_txty=txty_pred, pred_twth=twth_pred, label=target, num_classes=self.num_classes)
return total_loss