alex_bn_lee

导航

< 2025年3月 >
日 一 二 三 四 五 六
23 24 25 26 27 28 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 1 2 3 4 5

统计

【715】D-LinkNet实现

参考：遥感影像语义分割:D-LinkNet训练自己的数据集

参考：NekoApocalypse/road-extraction-d-linknet

代码：

?
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
from keras.layers import Conv2D, MaxPooling2D, Input, Conv2DTranspose, \
    Add, Activation, \
    BatchNormalization
from keras import Model
import keras
 
 
INPUT_SHAPE = (512, 512, 15)
 
 
def residual_block(input_tensor, num_filters):
    x = Conv2D(num_filters, (3, 3), padding='same')(input_tensor)
    x = Conv2D(num_filters, (3, 3), padding='same')(x)
    x = BatchNormalization()(x)
 
    input_tensor = Conv2D(num_filters, (1, 1), padding='same')(input_tensor)
    input_tensor = BatchNormalization()(input_tensor)
    res_tensor = Add()([input_tensor, x])
    res_tensor = Activation('relu')(res_tensor)
    return res_tensor
 
 
def dilated_center_block(input_tensor, num_filters):
 
    dilation_1 = Conv2D(num_filters, kernel_size=(3, 3), dilation_rate=(1, 1), padding='same')(input_tensor)
    dilation_1 = Activation('relu')(dilation_1)
 
    dilation_2 = Conv2D(num_filters, kernel_size=(3, 3), dilation_rate=(2, 2), padding='same')(dilation_1)
    dilation_2 = Activation('relu')(dilation_2)
 
    dilation_4 = Conv2D(num_filters, kernel_size=(3, 3), dilation_rate=(4, 4), padding='same')(dilation_2)
    dilation_4 = Activation('relu')(dilation_4)
 
    dilation_8 = Conv2D(num_filters, kernel_size=(3, 3), dilation_rate=(8, 8), padding='same')(dilation_4)
    dilation_8 = Activation('relu')(dilation_8)
 
    final_diliation = Add()([input_tensor, dilation_1, dilation_2, dilation_4, dilation_8])
 
    return final_diliation
 
 
def decoder_block(input_tensor, num_filters):
    decoder_tensor = Conv2D(num_filters, (1, 1), padding='same')(input_tensor)
    decoder_tensor = BatchNormalization()(decoder_tensor)
    decoder_tensor = Activation('relu')(decoder_tensor)
 
    decoder_tensor = Conv2DTranspose(num_filters, kernel_size=(3, 3), strides=(2, 2), padding='same')(decoder_tensor)
    decoder_tensor = BatchNormalization()(decoder_tensor)
    decoder_tensor = Activation('relu')(decoder_tensor)
 
    decoder_tensor = Conv2D(num_filters, (1, 1), padding='same')(decoder_tensor)
    decoder_tensor = BatchNormalization()(decoder_tensor)
    decoder_tensor = Activation('relu')(decoder_tensor)
    return decoder_tensor
 
 
def encoder_block(input_tensor, num_filters, num_res_blocks):
    encoded = residual_block(input_tensor, num_filters)
    while num_res_blocks > 1:
        encoded = residual_block(encoded, num_filters)
        num_res_blocks -= 1
    encoded_pool = MaxPooling2D((2, 2), strides=(2, 2))(encoded)
    return encoded, encoded_pool
 
 
def create_dlinknet():
    inputs = Input(shape=INPUT_SHAPE)
    inputs_ = Conv2D(64, kernel_size=(3, 3), padding='same')(inputs)
    inputs_ = BatchNormalization()(inputs_)
    inputs_ = Activation('relu')(inputs_)
    max_pool_inputs = MaxPooling2D((2, 2), strides=(2, 2))(inputs_)
 
    encoded_1, encoded_pool_1 = encoder_block(max_pool_inputs, num_filters=64, num_res_blocks=3)
    encoded_2, encoded_pool_2 = encoder_block(encoded_pool_1, num_filters=128, num_res_blocks=4)
    encoded_3, encoded_pool_3 = encoder_block(encoded_pool_2, num_filters=256, num_res_blocks=6)
    encoded_4, encoded_pool_4 = encoder_block(encoded_pool_3, num_filters=512, num_res_blocks=3)
 
    center = dilated_center_block(encoded_4, 512)
 
    decoded_1 = Add()([decoder_block(center, 256), encoded_3])
    decoded_2 = Add()([decoder_block(decoded_1, 128), encoded_2])
    decoded_3 = Add()([decoder_block(decoded_2, 64), encoded_1])
    decoded_4 = decoder_block(decoded_3, 64)
 
    final = Conv2DTranspose(32, kernel_size=(3, 3), padding='same')(decoded_4)
    outputs = Conv2D(1, (1, 1), activation='sigmoid')(final)
    model_i = Model(inputs=[inputs], outputs=[outputs])
    #model_i.compile(optimizer='adam', loss=combined_loss, metrics=[dice_coeff])
    #model_i.summary()
    # model_i.load_weights(save_model_path)
    return model_i
 
model = create_dlinknet()
model.summary()

输出：

Model: "model_1"
__________________________________________________________________________________________________
Layer (type)                    Output Shape         Param #     Connected to                     
==================================================================================================
input_2 (InputLayer)            [(None, 512, 512, 15 0                                            
__________________________________________________________________________________________________
conv2d_62 (Conv2D)              (None, 512, 512, 64) 8704        input_2[0][0]                    
__________________________________________________________________________________________________
batch_normalization_45 (BatchNo (None, 512, 512, 64) 256         conv2d_62[0][0]                  
__________________________________________________________________________________________________
activation_33 (Activation)      (None, 512, 512, 64) 0           batch_normalization_45[0][0]     
__________________________________________________________________________________________________
max_pooling2d_5 (MaxPooling2D)  (None, 256, 256, 64) 0           activation_33[0][0]              
__________________________________________________________________________________________________
conv2d_63 (Conv2D)              (None, 256, 256, 64) 36928       max_pooling2d_5[0][0]            
__________________________________________________________________________________________________
conv2d_65 (Conv2D)              (None, 256, 256, 64) 4160        max_pooling2d_5[0][0]            
__________________________________________________________________________________________________
conv2d_64 (Conv2D)              (None, 256, 256, 64) 36928       conv2d_63[0][0]                  
__________________________________________________________________________________________________
batch_normalization_47 (BatchNo (None, 256, 256, 64) 256         conv2d_65[0][0]                  
__________________________________________________________________________________________________
batch_normalization_46 (BatchNo (None, 256, 256, 64) 256         conv2d_64[0][0]                  
__________________________________________________________________________________________________
add_20 (Add)                    (None, 256, 256, 64) 0           batch_normalization_47[0][0]     
                                                                 batch_normalization_46[0][0]     
__________________________________________________________________________________________________
activation_34 (Activation)      (None, 256, 256, 64) 0           add_20[0][0]                     
__________________________________________________________________________________________________
conv2d_66 (Conv2D)              (None, 256, 256, 64) 36928       activation_34[0][0]              
__________________________________________________________________________________________________
conv2d_68 (Conv2D)              (None, 256, 256, 64) 4160        activation_34[0][0]              
__________________________________________________________________________________________________
conv2d_67 (Conv2D)              (None, 256, 256, 64) 36928       conv2d_66[0][0]                  
__________________________________________________________________________________________________
batch_normalization_49 (BatchNo (None, 256, 256, 64) 256         conv2d_68[0][0]                  
__________________________________________________________________________________________________
batch_normalization_48 (BatchNo (None, 256, 256, 64) 256         conv2d_67[0][0]                  
__________________________________________________________________________________________________
add_21 (Add)                    (None, 256, 256, 64) 0           batch_normalization_49[0][0]     
                                                                 batch_normalization_48[0][0]     
__________________________________________________________________________________________________
activation_35 (Activation)      (None, 256, 256, 64) 0           add_21[0][0]                     
__________________________________________________________________________________________________
conv2d_69 (Conv2D)              (None, 256, 256, 64) 36928       activation_35[0][0]              
__________________________________________________________________________________________________
conv2d_71 (Conv2D)              (None, 256, 256, 64) 4160        activation_35[0][0]              
__________________________________________________________________________________________________
conv2d_70 (Conv2D)              (None, 256, 256, 64) 36928       conv2d_69[0][0]                  
__________________________________________________________________________________________________
batch_normalization_51 (BatchNo (None, 256, 256, 64) 256         conv2d_71[0][0]                  
__________________________________________________________________________________________________
batch_normalization_50 (BatchNo (None, 256, 256, 64) 256         conv2d_70[0][0]                  
__________________________________________________________________________________________________
add_22 (Add)                    (None, 256, 256, 64) 0           batch_normalization_51[0][0]     
                                                                 batch_normalization_50[0][0]     
__________________________________________________________________________________________________
activation_36 (Activation)      (None, 256, 256, 64) 0           add_22[0][0]                     
__________________________________________________________________________________________________
max_pooling2d_6 (MaxPooling2D)  (None, 128, 128, 64) 0           activation_36[0][0]              
__________________________________________________________________________________________________
conv2d_72 (Conv2D)              (None, 128, 128, 128 73856       max_pooling2d_6[0][0]            
__________________________________________________________________________________________________
conv2d_74 (Conv2D)              (None, 128, 128, 128 8320        max_pooling2d_6[0][0]            
__________________________________________________________________________________________________
conv2d_73 (Conv2D)              (None, 128, 128, 128 147584      conv2d_72[0][0]                  
__________________________________________________________________________________________________
batch_normalization_53 (BatchNo (None, 128, 128, 128 512         conv2d_74[0][0]                  
__________________________________________________________________________________________________
batch_normalization_52 (BatchNo (None, 128, 128, 128 512         conv2d_73[0][0]                  
__________________________________________________________________________________________________
add_23 (Add)                    (None, 128, 128, 128 0           batch_normalization_53[0][0]     
                                                                 batch_normalization_52[0][0]     
__________________________________________________________________________________________________
activation_37 (Activation)      (None, 128, 128, 128 0           add_23[0][0]                     
__________________________________________________________________________________________________
conv2d_75 (Conv2D)              (None, 128, 128, 128 147584      activation_37[0][0]              
__________________________________________________________________________________________________
conv2d_77 (Conv2D)              (None, 128, 128, 128 16512       activation_37[0][0]              
__________________________________________________________________________________________________
conv2d_76 (Conv2D)              (None, 128, 128, 128 147584      conv2d_75[0][0]                  
__________________________________________________________________________________________________
batch_normalization_55 (BatchNo (None, 128, 128, 128 512         conv2d_77[0][0]                  
__________________________________________________________________________________________________
batch_normalization_54 (BatchNo (None, 128, 128, 128 512         conv2d_76[0][0]                  
__________________________________________________________________________________________________
add_24 (Add)                    (None, 128, 128, 128 0           batch_normalization_55[0][0]     
                                                                 batch_normalization_54[0][0]     
__________________________________________________________________________________________________
activation_38 (Activation)      (None, 128, 128, 128 0           add_24[0][0]                     
__________________________________________________________________________________________________
conv2d_78 (Conv2D)              (None, 128, 128, 128 147584      activation_38[0][0]              
__________________________________________________________________________________________________
conv2d_80 (Conv2D)              (None, 128, 128, 128 16512       activation_38[0][0]              
__________________________________________________________________________________________________
conv2d_79 (Conv2D)              (None, 128, 128, 128 147584      conv2d_78[0][0]                  
__________________________________________________________________________________________________
batch_normalization_57 (BatchNo (None, 128, 128, 128 512         conv2d_80[0][0]                  
__________________________________________________________________________________________________
batch_normalization_56 (BatchNo (None, 128, 128, 128 512         conv2d_79[0][0]                  
__________________________________________________________________________________________________
add_25 (Add)                    (None, 128, 128, 128 0           batch_normalization_57[0][0]     
                                                                 batch_normalization_56[0][0]     
__________________________________________________________________________________________________
activation_39 (Activation)      (None, 128, 128, 128 0           add_25[0][0]                     
__________________________________________________________________________________________________
conv2d_81 (Conv2D)              (None, 128, 128, 128 147584      activation_39[0][0]              
__________________________________________________________________________________________________
conv2d_83 (Conv2D)              (None, 128, 128, 128 16512       activation_39[0][0]              
__________________________________________________________________________________________________
conv2d_82 (Conv2D)              (None, 128, 128, 128 147584      conv2d_81[0][0]                  
__________________________________________________________________________________________________
batch_normalization_59 (BatchNo (None, 128, 128, 128 512         conv2d_83[0][0]                  
__________________________________________________________________________________________________
batch_normalization_58 (BatchNo (None, 128, 128, 128 512         conv2d_82[0][0]                  
__________________________________________________________________________________________________
add_26 (Add)                    (None, 128, 128, 128 0           batch_normalization_59[0][0]     
                                                                 batch_normalization_58[0][0]     
__________________________________________________________________________________________________
activation_40 (Activation)      (None, 128, 128, 128 0           add_26[0][0]                     
__________________________________________________________________________________________________
max_pooling2d_7 (MaxPooling2D)  (None, 64, 64, 128)  0           activation_40[0][0]              
__________________________________________________________________________________________________
conv2d_84 (Conv2D)              (None, 64, 64, 256)  295168      max_pooling2d_7[0][0]            
__________________________________________________________________________________________________
conv2d_86 (Conv2D)              (None, 64, 64, 256)  33024       max_pooling2d_7[0][0]            
__________________________________________________________________________________________________
conv2d_85 (Conv2D)              (None, 64, 64, 256)  590080      conv2d_84[0][0]                  
__________________________________________________________________________________________________
batch_normalization_61 (BatchNo (None, 64, 64, 256)  1024        conv2d_86[0][0]                  
__________________________________________________________________________________________________
batch_normalization_60 (BatchNo (None, 64, 64, 256)  1024        conv2d_85[0][0]                  
__________________________________________________________________________________________________
add_27 (Add)                    (None, 64, 64, 256)  0           batch_normalization_61[0][0]     
                                                                 batch_normalization_60[0][0]     
__________________________________________________________________________________________________
activation_41 (Activation)      (None, 64, 64, 256)  0           add_27[0][0]                     
__________________________________________________________________________________________________
conv2d_87 (Conv2D)              (None, 64, 64, 256)  590080      activation_41[0][0]              
__________________________________________________________________________________________________
conv2d_89 (Conv2D)              (None, 64, 64, 256)  65792       activation_41[0][0]              
__________________________________________________________________________________________________
conv2d_88 (Conv2D)              (None, 64, 64, 256)  590080      conv2d_87[0][0]                  
__________________________________________________________________________________________________
batch_normalization_63 (BatchNo (None, 64, 64, 256)  1024        conv2d_89[0][0]                  
__________________________________________________________________________________________________
batch_normalization_62 (BatchNo (None, 64, 64, 256)  1024        conv2d_88[0][0]                  
__________________________________________________________________________________________________
add_28 (Add)                    (None, 64, 64, 256)  0           batch_normalization_63[0][0]     
                                                                 batch_normalization_62[0][0]     
__________________________________________________________________________________________________
activation_42 (Activation)      (None, 64, 64, 256)  0           add_28[0][0]                     
__________________________________________________________________________________________________
conv2d_90 (Conv2D)              (None, 64, 64, 256)  590080      activation_42[0][0]              
__________________________________________________________________________________________________
conv2d_92 (Conv2D)              (None, 64, 64, 256)  65792       activation_42[0][0]              
__________________________________________________________________________________________________
conv2d_91 (Conv2D)              (None, 64, 64, 256)  590080      conv2d_90[0][0]                  
__________________________________________________________________________________________________
batch_normalization_65 (BatchNo (None, 64, 64, 256)  1024        conv2d_92[0][0]                  
__________________________________________________________________________________________________
batch_normalization_64 (BatchNo (None, 64, 64, 256)  1024        conv2d_91[0][0]                  
__________________________________________________________________________________________________
add_29 (Add)                    (None, 64, 64, 256)  0           batch_normalization_65[0][0]     
                                                                 batch_normalization_64[0][0]     
__________________________________________________________________________________________________
activation_43 (Activation)      (None, 64, 64, 256)  0           add_29[0][0]                     
__________________________________________________________________________________________________
conv2d_93 (Conv2D)              (None, 64, 64, 256)  590080      activation_43[0][0]              
__________________________________________________________________________________________________
conv2d_95 (Conv2D)              (None, 64, 64, 256)  65792       activation_43[0][0]              
__________________________________________________________________________________________________
conv2d_94 (Conv2D)              (None, 64, 64, 256)  590080      conv2d_93[0][0]                  
__________________________________________________________________________________________________
batch_normalization_67 (BatchNo (None, 64, 64, 256)  1024        conv2d_95[0][0]                  
__________________________________________________________________________________________________
batch_normalization_66 (BatchNo (None, 64, 64, 256)  1024        conv2d_94[0][0]                  
__________________________________________________________________________________________________
add_30 (Add)                    (None, 64, 64, 256)  0           batch_normalization_67[0][0]     
                                                                 batch_normalization_66[0][0]     
__________________________________________________________________________________________________
activation_44 (Activation)      (None, 64, 64, 256)  0           add_30[0][0]                     
__________________________________________________________________________________________________
conv2d_96 (Conv2D)              (None, 64, 64, 256)  590080      activation_44[0][0]              
__________________________________________________________________________________________________
conv2d_98 (Conv2D)              (None, 64, 64, 256)  65792       activation_44[0][0]              
__________________________________________________________________________________________________
conv2d_97 (Conv2D)              (None, 64, 64, 256)  590080      conv2d_96[0][0]                  
__________________________________________________________________________________________________
batch_normalization_69 (BatchNo (None, 64, 64, 256)  1024        conv2d_98[0][0]                  
__________________________________________________________________________________________________
batch_normalization_68 (BatchNo (None, 64, 64, 256)  1024        conv2d_97[0][0]                  
__________________________________________________________________________________________________
add_31 (Add)                    (None, 64, 64, 256)  0           batch_normalization_69[0][0]     
                                                                 batch_normalization_68[0][0]     
__________________________________________________________________________________________________
activation_45 (Activation)      (None, 64, 64, 256)  0           add_31[0][0]                     
__________________________________________________________________________________________________
conv2d_99 (Conv2D)              (None, 64, 64, 256)  590080      activation_45[0][0]              
__________________________________________________________________________________________________
conv2d_101 (Conv2D)             (None, 64, 64, 256)  65792       activation_45[0][0]              
__________________________________________________________________________________________________
conv2d_100 (Conv2D)             (None, 64, 64, 256)  590080      conv2d_99[0][0]                  
__________________________________________________________________________________________________
batch_normalization_71 (BatchNo (None, 64, 64, 256)  1024        conv2d_101[0][0]                 
__________________________________________________________________________________________________
batch_normalization_70 (BatchNo (None, 64, 64, 256)  1024        conv2d_100[0][0]                 
__________________________________________________________________________________________________
add_32 (Add)                    (None, 64, 64, 256)  0           batch_normalization_71[0][0]     
                                                                 batch_normalization_70[0][0]     
__________________________________________________________________________________________________
activation_46 (Activation)      (None, 64, 64, 256)  0           add_32[0][0]                     
__________________________________________________________________________________________________
max_pooling2d_8 (MaxPooling2D)  (None, 32, 32, 256)  0           activation_46[0][0]              
__________________________________________________________________________________________________
conv2d_102 (Conv2D)             (None, 32, 32, 512)  1180160     max_pooling2d_8[0][0]            
__________________________________________________________________________________________________
conv2d_104 (Conv2D)             (None, 32, 32, 512)  131584      max_pooling2d_8[0][0]            
__________________________________________________________________________________________________
conv2d_103 (Conv2D)             (None, 32, 32, 512)  2359808     conv2d_102[0][0]                 
__________________________________________________________________________________________________
batch_normalization_73 (BatchNo (None, 32, 32, 512)  2048        conv2d_104[0][0]                 
__________________________________________________________________________________________________
batch_normalization_72 (BatchNo (None, 32, 32, 512)  2048        conv2d_103[0][0]                 
__________________________________________________________________________________________________
add_33 (Add)                    (None, 32, 32, 512)  0           batch_normalization_73[0][0]     
                                                                 batch_normalization_72[0][0]     
__________________________________________________________________________________________________
activation_47 (Activation)      (None, 32, 32, 512)  0           add_33[0][0]                     
__________________________________________________________________________________________________
conv2d_105 (Conv2D)             (None, 32, 32, 512)  2359808     activation_47[0][0]              
__________________________________________________________________________________________________
conv2d_107 (Conv2D)             (None, 32, 32, 512)  262656      activation_47[0][0]              
__________________________________________________________________________________________________
conv2d_106 (Conv2D)             (None, 32, 32, 512)  2359808     conv2d_105[0][0]                 
__________________________________________________________________________________________________
batch_normalization_75 (BatchNo (None, 32, 32, 512)  2048        conv2d_107[0][0]                 
__________________________________________________________________________________________________
batch_normalization_74 (BatchNo (None, 32, 32, 512)  2048        conv2d_106[0][0]                 
__________________________________________________________________________________________________
add_34 (Add)                    (None, 32, 32, 512)  0           batch_normalization_75[0][0]     
                                                                 batch_normalization_74[0][0]     
__________________________________________________________________________________________________
activation_48 (Activation)      (None, 32, 32, 512)  0           add_34[0][0]                     
__________________________________________________________________________________________________
conv2d_108 (Conv2D)             (None, 32, 32, 512)  2359808     activation_48[0][0]              
__________________________________________________________________________________________________
conv2d_110 (Conv2D)             (None, 32, 32, 512)  262656      activation_48[0][0]              
__________________________________________________________________________________________________
conv2d_109 (Conv2D)             (None, 32, 32, 512)  2359808     conv2d_108[0][0]                 
__________________________________________________________________________________________________
batch_normalization_77 (BatchNo (None, 32, 32, 512)  2048        conv2d_110[0][0]                 
__________________________________________________________________________________________________
batch_normalization_76 (BatchNo (None, 32, 32, 512)  2048        conv2d_109[0][0]                 
__________________________________________________________________________________________________
add_35 (Add)                    (None, 32, 32, 512)  0           batch_normalization_77[0][0]     
                                                                 batch_normalization_76[0][0]     
__________________________________________________________________________________________________
activation_49 (Activation)      (None, 32, 32, 512)  0           add_35[0][0]                     
__________________________________________________________________________________________________
conv2d_111 (Conv2D)             (None, 32, 32, 512)  2359808     activation_49[0][0]              
__________________________________________________________________________________________________
activation_50 (Activation)      (None, 32, 32, 512)  0           conv2d_111[0][0]                 
__________________________________________________________________________________________________
conv2d_112 (Conv2D)             (None, 32, 32, 512)  2359808     activation_50[0][0]              
__________________________________________________________________________________________________
activation_51 (Activation)      (None, 32, 32, 512)  0           conv2d_112[0][0]                 
__________________________________________________________________________________________________
conv2d_113 (Conv2D)             (None, 32, 32, 512)  2359808     activation_51[0][0]              
__________________________________________________________________________________________________
activation_52 (Activation)      (None, 32, 32, 512)  0           conv2d_113[0][0]                 
__________________________________________________________________________________________________
conv2d_114 (Conv2D)             (None, 32, 32, 512)  2359808     activation_52[0][0]              
__________________________________________________________________________________________________
activation_53 (Activation)      (None, 32, 32, 512)  0           conv2d_114[0][0]                 
__________________________________________________________________________________________________
add_36 (Add)                    (None, 32, 32, 512)  0           activation_49[0][0]              
                                                                 activation_50[0][0]              
                                                                 activation_51[0][0]              
                                                                 activation_52[0][0]              
                                                                 activation_53[0][0]              
__________________________________________________________________________________________________
conv2d_115 (Conv2D)             (None, 32, 32, 256)  131328      add_36[0][0]                     
__________________________________________________________________________________________________
batch_normalization_78 (BatchNo (None, 32, 32, 256)  1024        conv2d_115[0][0]                 
__________________________________________________________________________________________________
activation_54 (Activation)      (None, 32, 32, 256)  0           batch_normalization_78[0][0]     
__________________________________________________________________________________________________
conv2d_transpose_5 (Conv2DTrans (None, 64, 64, 256)  590080      activation_54[0][0]              
__________________________________________________________________________________________________
batch_normalization_79 (BatchNo (None, 64, 64, 256)  1024        conv2d_transpose_5[0][0]         
__________________________________________________________________________________________________
activation_55 (Activation)      (None, 64, 64, 256)  0           batch_normalization_79[0][0]     
__________________________________________________________________________________________________
conv2d_116 (Conv2D)             (None, 64, 64, 256)  65792       activation_55[0][0]              
__________________________________________________________________________________________________
batch_normalization_80 (BatchNo (None, 64, 64, 256)  1024        conv2d_116[0][0]                 
__________________________________________________________________________________________________
activation_56 (Activation)      (None, 64, 64, 256)  0           batch_normalization_80[0][0]     
__________________________________________________________________________________________________
add_37 (Add)                    (None, 64, 64, 256)  0           activation_56[0][0]              
                                                                 activation_46[0][0]              
__________________________________________________________________________________________________
conv2d_117 (Conv2D)             (None, 64, 64, 128)  32896       add_37[0][0]                     
__________________________________________________________________________________________________
batch_normalization_81 (BatchNo (None, 64, 64, 128)  512         conv2d_117[0][0]                 
__________________________________________________________________________________________________
activation_57 (Activation)      (None, 64, 64, 128)  0           batch_normalization_81[0][0]     
__________________________________________________________________________________________________
conv2d_transpose_6 (Conv2DTrans (None, 128, 128, 128 147584      activation_57[0][0]              
__________________________________________________________________________________________________
batch_normalization_82 (BatchNo (None, 128, 128, 128 512         conv2d_transpose_6[0][0]         
__________________________________________________________________________________________________
activation_58 (Activation)      (None, 128, 128, 128 0           batch_normalization_82[0][0]     
__________________________________________________________________________________________________
conv2d_118 (Conv2D)             (None, 128, 128, 128 16512       activation_58[0][0]              
__________________________________________________________________________________________________
batch_normalization_83 (BatchNo (None, 128, 128, 128 512         conv2d_118[0][0]                 
__________________________________________________________________________________________________
activation_59 (Activation)      (None, 128, 128, 128 0           batch_normalization_83[0][0]     
__________________________________________________________________________________________________
add_38 (Add)                    (None, 128, 128, 128 0           activation_59[0][0]              
                                                                 activation_40[0][0]              
__________________________________________________________________________________________________
conv2d_119 (Conv2D)             (None, 128, 128, 64) 8256        add_38[0][0]                     
__________________________________________________________________________________________________
batch_normalization_84 (BatchNo (None, 128, 128, 64) 256         conv2d_119[0][0]                 
__________________________________________________________________________________________________
activation_60 (Activation)      (None, 128, 128, 64) 0           batch_normalization_84[0][0]     
__________________________________________________________________________________________________
conv2d_transpose_7 (Conv2DTrans (None, 256, 256, 64) 36928       activation_60[0][0]              
__________________________________________________________________________________________________
batch_normalization_85 (BatchNo (None, 256, 256, 64) 256         conv2d_transpose_7[0][0]         
__________________________________________________________________________________________________
activation_61 (Activation)      (None, 256, 256, 64) 0           batch_normalization_85[0][0]     
__________________________________________________________________________________________________
conv2d_120 (Conv2D)             (None, 256, 256, 64) 4160        activation_61[0][0]              
__________________________________________________________________________________________________
batch_normalization_86 (BatchNo (None, 256, 256, 64) 256         conv2d_120[0][0]                 
__________________________________________________________________________________________________
activation_62 (Activation)      (None, 256, 256, 64) 0           batch_normalization_86[0][0]     
__________________________________________________________________________________________________
add_39 (Add)                    (None, 256, 256, 64) 0           activation_62[0][0]              
                                                                 activation_36[0][0]              
__________________________________________________________________________________________________
conv2d_121 (Conv2D)             (None, 256, 256, 64) 4160        add_39[0][0]                     
__________________________________________________________________________________________________
batch_normalization_87 (BatchNo (None, 256, 256, 64) 256         conv2d_121[0][0]                 
__________________________________________________________________________________________________
activation_63 (Activation)      (None, 256, 256, 64) 0           batch_normalization_87[0][0]     
__________________________________________________________________________________________________
conv2d_transpose_8 (Conv2DTrans (None, 512, 512, 64) 36928       activation_63[0][0]              
__________________________________________________________________________________________________
batch_normalization_88 (BatchNo (None, 512, 512, 64) 256         conv2d_transpose_8[0][0]         
__________________________________________________________________________________________________
activation_64 (Activation)      (None, 512, 512, 64) 0           batch_normalization_88[0][0]     
__________________________________________________________________________________________________
conv2d_122 (Conv2D)             (None, 512, 512, 64) 4160        activation_64[0][0]              
__________________________________________________________________________________________________
batch_normalization_89 (BatchNo (None, 512, 512, 64) 256         conv2d_122[0][0]                 
__________________________________________________________________________________________________
activation_65 (Activation)      (None, 512, 512, 64) 0           batch_normalization_89[0][0]     
__________________________________________________________________________________________________
conv2d_transpose_9 (Conv2DTrans (None, 512, 512, 32) 18464       activation_65[0][0]              
__________________________________________________________________________________________________
conv2d_123 (Conv2D)             (None, 512, 512, 1)  33          conv2d_transpose_9[0][0]         
==================================================================================================
Total params: 32,764,801
Trainable params: 32,746,497
Non-trainable params: 18,304
__________________________________________________________________________________________________

结构图： 

 

posted on   McDelfino  阅读(287)  评论(0)  编辑  收藏  举报

相关博文：
· 【714】LinkNet实现
· 【706】Keras官网语义分割例子解读
· 深度学习之残差网络
· ShuffleNetV1:极致轻量化卷积神经网络(分组卷积+通道重排)
· 别玩手机 图像分类比赛
阅读排行：
· DeepSeek 开源周回顾「GitHub 热点速览」
· 记一次.NET内存居高不下排查解决与启示
· 物流快递公司核心技术能力-地址解析分单基础技术分享
· .NET 10首个预览版发布：重大改进与新特性概览！
· .NET10 - 预览版1新功能体验（一）
历史上的今天：
2021-05-31 【569】LaTex, arrows over letters
2021-05-31 【568】review
2021-05-31 【567】Convert LaTex to Word Using Pandoc
2012-05-31 【043】光盘库编码
点击右上角即可分享
微信分享提示
AI FOR CODE 大赛