Skip to content

Releases: yayoimizuha/helloproject-blog-image-clawler

First Release!

12 Jan 05:18
Compare
Choose a tag to compare

Hello!Projectのメンバーのアメブロから収集した画像を、MediaPipeのBlazeFaceで顔検出&切り出しを行い、VGGFace2で学習させたFaceNetを用いて特徴量を抽出して、マンハッタン距離でtopKアルゴリズムで分類したものです。

2023-01-15 12:13:58.165751.h5の中身(model.summary())
__________________________________________________________________________________________________
 Layer (type)                   Output Shape         Param #     Connected to                     
==================================================================================================
 input_1 (InputLayer)           [(None, 224, 224, 3  0           []                               
                                )]                                                                
                                                                                                  
 conv1/7x7_s2 (Conv2D)          (None, 112, 112, 64  9408        ['input_1[0][0]']                
                                )                                                                 
                                                                                                  
 conv1/7x7_s2/bn (BatchNormaliz  (None, 112, 112, 64  256        ['conv1/7x7_s2[0][0]']           
 ation)                         )                                                                 
                                                                                                  
 activation (Activation)        (None, 112, 112, 64  0           ['conv1/7x7_s2/bn[0][0]']        
                                )                                                                 
                                                                                                  
 max_pooling2d (MaxPooling2D)   (None, 55, 55, 64)   0           ['activation[0][0]']             
                                                                                                  
 conv2_1_1x1_reduce (Conv2D)    (None, 55, 55, 64)   4096        ['max_pooling2d[0][0]']          
                                                                                                  
 conv2_1_1x1_reduce/bn (BatchNo  (None, 55, 55, 64)  256         ['conv2_1_1x1_reduce[0][0]']     
 rmalization)                                                                                     
                                                                                                  
 activation_1 (Activation)      (None, 55, 55, 64)   0           ['conv2_1_1x1_reduce/bn[0][0]']  
                                                                                                  
 conv2_1_3x3 (Conv2D)           (None, 55, 55, 64)   36864       ['activation_1[0][0]']           
                                                                                                  
 conv2_1_3x3/bn (BatchNormaliza  (None, 55, 55, 64)  256         ['conv2_1_3x3[0][0]']            
 tion)                                                                                            
                                                                                                  
 activation_2 (Activation)      (None, 55, 55, 64)   0           ['conv2_1_3x3/bn[0][0]']         
                                                                                                  
 conv2_1_1x1_increase (Conv2D)  (None, 55, 55, 256)  16384       ['activation_2[0][0]']           
                                                                                                  
 conv2_1_1x1_proj (Conv2D)      (None, 55, 55, 256)  16384       ['max_pooling2d[0][0]']          
                                                                                                  
 conv2_1_1x1_increase/bn (Batch  (None, 55, 55, 256)  1024       ['conv2_1_1x1_increase[0][0]']   
 Normalization)                                                                                   
                                                                                                  
 conv2_1_1x1_proj/bn (BatchNorm  (None, 55, 55, 256)  1024       ['conv2_1_1x1_proj[0][0]']       
 alization)                                                                                       
                                                                                                  
 add (Add)                      (None, 55, 55, 256)  0           ['conv2_1_1x1_increase/bn[0][0]',
                                                                  'conv2_1_1x1_proj/bn[0][0]']    
                                                                                                  
 activation_3 (Activation)      (None, 55, 55, 256)  0           ['add[0][0]']                    
                                                                                                  
 conv2_2_1x1_reduce (Conv2D)    (None, 55, 55, 64)   16384       ['activation_3[0][0]']           
                                                                                                  
 conv2_2_1x1_reduce/bn (BatchNo  (None, 55, 55, 64)  256         ['conv2_2_1x1_reduce[0][0]']     
 rmalization)                                                                                     
                                                                                                  
 activation_4 (Activation)      (None, 55, 55, 64)   0           ['conv2_2_1x1_reduce/bn[0][0]']  
                                                                                                  
 conv2_2_3x3 (Conv2D)           (None, 55, 55, 64)   36864       ['activation_4[0][0]']           
                                                                                                  
 conv2_2_3x3/bn (BatchNormaliza  (None, 55, 55, 64)  256         ['conv2_2_3x3[0][0]']            
 tion)                                                                                            
                                                                                                  
 activation_5 (Activation)      (None, 55, 55, 64)   0           ['conv2_2_3x3/bn[0][0]']         
                                                                                                  
 conv2_2_1x1_increase (Conv2D)  (None, 55, 55, 256)  16384       ['activation_5[0][0]']           
                                                                                                  
 conv2_2_1x1_increase/bn (Batch  (None, 55, 55, 256)  1024       ['conv2_2_1x1_increase[0][0]']   
 Normalization)                                                                                   
                                                                                                  
 add_1 (Add)                    (None, 55, 55, 256)  0           ['conv2_2_1x1_increase/bn[0][0]',
                                                                  'activation_3[0][0]']           
                                                                                                  
 activation_6 (Activation)      (None, 55, 55, 256)  0           ['add_1[0][0]']                  
                                                                                                  
 conv2_3_1x1_reduce (Conv2D)    (None, 55, 55, 64)   16384       ['activation_6[0][0]']           
                                                                                                  
 conv2_3_1x1_reduce/bn (BatchNo  (None, 55, 55, 64)  256         ['conv2_3_1x1_reduce[0][0]']     
 rmalization)                                                                                     
                                                                                                  
 activation_7 (Activation)      (None, 55, 55, 64)   0           ['conv2_3_1x1_reduce/bn[0][0]']  
                                                                                                  
 conv2_3_3x3 (Conv2D)           (None, 55, 55, 64)   36864       ['activation_7[0][0]']           
                                                                                                  
 conv2_3_3x3/bn (BatchNormaliza  (None, 55, 55, 64)  256         ['conv2_3_3x3[0][0]']            
 tion)                                                                                            
                                                                                                  
 activation_8 (Activation)      (None, 55, 55, 64)   0           ['conv2_3_3x3/bn[0][0]']         
                                                                                                  
 conv2_3_1x1_increase (Conv2D)  (None, 55, 55, 256)  16384       ['activation_8[0][0]']           
                                                                                                  
 conv2_3_1x1_increase/bn (Batch  (None, 55, 55, 256)  1024       ['conv2_3_1x1_increase[0][0]']   
 Normalization)                                                                                   
                                                                                                  
 add_2 (Add)                    (None, 55, 55, 256)  0           ['conv2_3_1x1_increase/bn[0][0]',
                                                                  'activation_6[0][0]']           
                                                                                                  
 activation_9 (Activation)      (None, 55, 55, 256)  0           ['add_2[0][0]']                  
                                                                                                  
 conv3_1_1x1_reduce (Conv2D)    (None, 28, 28, 128)  32768       ['activation_9[0][0]']           
                                                                                                  
 conv3_1_1x1_reduce/bn (BatchNo  (None, 28, 28, 128)  512        ['conv3_1_1x1_reduce[0][0]']     
 rmalization)                                                                ...
Read more