-
Notifications
You must be signed in to change notification settings - Fork 0
/
sfp_resnet50.m
178 lines (157 loc) · 5.67 KB
/
sfp_resnet50.m
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
%%%%%%%%%%%%%%%%%%%%%%%input(224*224*3)%%%%%%%%%%%%%%%%%%%%%%%
clear;
original_pic = imread('ILSVRC2012_val_00000018.JPEG');
original_pic = single(original_pic); %kenengmeiyouwenti
modelfile = 'resnet_50.h5';
lgraph = importKerasLayers(modelfile,'ImportWeights',true);
eps = 1e-5;
%original image:float32_to_sfp
%conv1_zeropadding(230*230*3),(3,3,3,3)
image_r = original_pic(:,:,1);
image_g = original_pic(:,:,2);
image_b = original_pic(:,:,3);
image_r = padarray(image_r, [3 3]);
image_g = padarray(image_g, [3 3]);
image_b = padarray(image_b, [3 3]);
% image_r = padarray(image_r, [1 1]);%padding[0,0,0,0]
% image_g = padarray(image_g, [1 1]);
% image_b = padarray(image_b, [1 1]); %[232*232]
% r=image_r([3:9],[3:9]);
% g=image_g([3:9],[3:9]);
% b=image_b([3:9],[3:9]);
%merge
weights = lgraph.Layers(3,1).Weights; %7*7*3*64
bias1 = lgraph.Layers(3,1).Bias; %1*1*64
size_weight = size(weights); %[7 7 3 64]
image_channels = size_weight(3);
filters = size_weight(4); %64
kernels = size_weight(1); %7
trained_mean = lgraph.Layers(4,1).TrainedMean;
trained_variance = lgraph.Layers(4,1).TrainedVariance;
beta = lgraph.Layers(4,1).Offset;
gamma = lgraph.Layers(4,1).Scale; %[1*1*64]
for weight_filters = 1:filters
for weight_channels = 1:image_channels
for weight_row = 1:kernels
for weight_line = 1:kernels
weights(weight_row,weight_line,weight_channels,weight_filters) = weights(weight_row,weight_line,weight_channels,weight_filters) * gamma(:,:,weight_filters) / (sqrt(trained_variance(:,:,weight_filters) + eps));
end
end
end
bias1(:,:,weight_filters) = gamma(:,:,weight_filters) * (bias1(:,:,weight_filters) - trained_mean(:,:,weight_filters)) / (sqrt(trained_variance(:,:,weight_filters) + eps)) + beta(:,:,weight_filters);
end
%
% weights(:,:,1,1);
% ba=bias1(:,:,1);
% r=r.*weights(:,:,1,1);
% g=g.*weights(:,:,2,1);
% b=b.*weights(:,:,3,1);
% sum(sum(r));
% sum(sum(g));
% sum(sum(b));
% s=sum(sum(r))+sum(sum(b))+sum(sum(g))+ba;
%conv1_conv
image_r = padarray(image_r, [1 1]);%padding[0,0,0,0]
image_g = padarray(image_g, [1 1]);
image_b = padarray(image_b, [1 1]); %[232*232]
pic(:,:,1) = image_r;
pic(:,:,2) = image_g;
pic(:,:,3) = image_b;
image_size = size(image_r);
image_size = image_size(1);
strides = 2;
image_conv_row = (image_size - kernels + 1) / strides;
image_conv_line = ceil(image_conv_row);
image_conv_row = image_conv_line;
for filter_conv = 1:filters
for image_conv_channels = 1:image_channels
for image_conv_r = 1:image_conv_row
for image_conv_l = 1:image_conv_line
conv_mul = [];
count = 1;
pic_r = 1 + strides * (image_conv_r - 1) ;
for conv_r = 1:kernels
pic_l = 1 + strides * (image_conv_l - 1);
for conv_l = 1:kernels
conv_mul(count) = pic(pic_r,pic_l,image_conv_channels) * weights(conv_r,conv_l,image_conv_channels,filter_conv);
if pic_l < image_size
pic_l = pic_l + 1;
count = count + 1;
end
end
pic_r = pic_r + 1;
end
image_conv(image_conv_r,image_conv_l,image_conv_channels,filter_conv) = sum(conv_mul);
end
end
end
end
% image_conv([1:4],[1:4],1,1);
% image_conv(:,:,2,1);
% image_conv(:,:,3,1);
image_conv_out = zeros(image_conv_row,image_conv_line,filters);
for filter_conv = 1:filters
for image_conv_r = 1:image_conv_row
for image_conv_l = 1:image_conv_line
for image_conv_channels = 1:image_channels
image_conv_out(image_conv_r,image_conv_l,filter_conv) = image_conv(image_conv_r,image_conv_l,image_conv_channels,filter_conv) + image_conv_out(image_conv_r,image_conv_l,filter_conv);
end
image_conv_out(image_conv_r,image_conv_l,filter_conv) = image_conv_out(image_conv_r,image_conv_l,filter_conv) + bias1(:,:,filter_conv);
end
end
end
image_conv_out(:,:,1);
%conv1_relu
for filter_bn = 1:filters
for image_bn_r = 1:image_conv_row
for image_bn_l = 1:image_conv_line
if image_conv_out(image_bn_r,image_bn_l,filter_bn) < 0
image_conv_out(image_bn_r,image_bn_l,filter_bn) = 0;
end
end
end
end
image_conv_out(:,:,1);
%ZeroPadding(1,1,1,1)
for filter_padding = 1:filters
image_padding = image_conv_out(:,:,filter_padding);
image_padding = padarray(image_padding, [1 1]);
image_padding_out(:,:,filter_padding) = image_padding;
end
%MaxPooling(3*3,strides=2)
strides = 2;
pool_size = 3;
for filter_pool = 1:filters
image_pool = image_padding_out(:,:,filter_pool);
image_pool = padarray(image_pool, [1 1]);
image_pool_out(:,:,filter_pool) = image_pool; %padding(0,0,0,0)
end
image_pool_out(:,:,1);
image_size = size(image_pool_out);
image_size = image_size(1);
image_pool_row = (image_size - pool_size + 1) / 2;
image_pool_line = ceil(image_pool_row);
image_pool_row = image_pool_line;
image_pool = zeros(image_pool_row,image_pool_line,filters);
for filter_pool = 1:filters
for image_pool_r = 1:image_pool_row
for image_pool_l = 1:image_pool_line
image_r = 1 + strides * (image_pool_r - 1);
for pool_r = 1:pool_size
image_l = 1 + strides * (image_pool_l - 1);
for pool_l = 1:pool_size
if image_pool(image_pool_r,image_pool_l,filter_pool) < image_pool_out(image_r,image_l,filter_pool)
image_pool(image_pool_r,image_pool_l,filter_pool) = image_pool_out(image_r,image_l,filter_pool);
end
if image_r < image_size
image_l = image_l + 1;
end
end
if image_l <image_size
image_r = image_r + 1;
end
end
end
end
end
image_pool([1:3],[1:3],1)