質問編集履歴
1
質問内容の修正
test
CHANGED
|
File without changes
|
test
CHANGED
|
@@ -46,7 +46,21 @@
|
|
|
46
46
|
|
|
47
47
|
```
|
|
48
48
|
|
|
49
|
-
|
|
49
|
+
```
|
|
50
|
+
|
|
51
|
+
layer_outputs = [layer.output for layer in model.layers[1:2]]
|
|
52
|
+
|
|
53
|
+
activation_model = Model(inputs=model.input, outputs=layer_outputs)
|
|
54
|
+
|
|
55
|
+
activation_model.summary()
|
|
56
|
+
|
|
57
|
+
|
|
58
|
+
|
|
59
|
+
|
|
60
|
+
|
|
61
|
+
コード
|
|
62
|
+
|
|
63
|
+
```
|
|
50
64
|
|
|
51
65
|
###
|
|
52
66
|
|
|
@@ -56,15 +70,15 @@
|
|
|
56
70
|
|
|
57
71
|
畳み込みフィルタに関しては，2個目のコードで可視化を行えました。
|
|
58
72
|
|
|
59
|
-
質問したい点は2点です。
|
|
60
|
-
|
|
61
|
-
1.入力画像はグレースケール画像を用いているのですが，フィルタを可視化してみるとカラー画像でした。
|
|
62
|
-
|
|
63
|
-
自分のイメージとしては，入力画像がグレースケールであれば，フィルタ画像もグレースケールなように感じているのですが，違うのでしょうか，，，
|
|
64
|
-
|
|
65
|
-
|
|
73
|
+
特徴マップを可視化する方法をお聞きしたいのですが，
|
|
74
|
+
|
|
66
|
-
|
|
75
|
+
3個目のコードで中間層を可視化するモデル（今はconv2D層のみ）を作成しようとしたのですが，下記エラーが発生しました。
|
|
76
|
+
|
|
67
|
-
|
|
77
|
+
どなたかよろしくお願い致します。
|
|
78
|
+
|
|
79
|
+
|
|
80
|
+
|
|
81
|
+
|
|
68
82
|
|
|
69
83
|
### 発生している問題・エラーメッセージ
|
|
70
84
|
|
|
@@ -72,59 +86,93 @@
|
|
|
72
86
|
|
|
73
87
|
```
|
|
74
88
|
|
|
89
|
+
TypeError Traceback (most recent call last)
|
|
90
|
+
|
|
91
|
+
<ipython-input-55-1659aa1308fe> in <module>
|
|
92
|
+
|
|
93
|
+
----> 1 activation_model = Model(inputs=model.input, outputs=layer_outputs)
|
|
94
|
+
|
|
95
|
+
2 activation_model.summary()
|
|
96
|
+
|
|
97
|
+
|
|
98
|
+
|
|
99
|
+
~\Anaconda3\lib\site-packages\keras\legacy\interfaces.py in wrapper(*args, **kwargs)
|
|
100
|
+
|
|
101
|
+
89 warnings.warn('Update your `' + object_name + '` call to the ' +
|
|
102
|
+
|
|
103
|
+
90 'Keras 2 API: ' + signature, stacklevel=2)
|
|
104
|
+
|
|
105
|
+
---> 91 return func(*args, **kwargs)
|
|
106
|
+
|
|
107
|
+
92 wrapper._original_function = func
|
|
108
|
+
|
|
109
|
+
93 return wrapper
|
|
110
|
+
|
|
111
|
+
|
|
112
|
+
|
|
113
|
+
~\Anaconda3\lib\site-packages\keras\engine\network.py in __init__(self, *args, **kwargs)
|
|
114
|
+
|
|
115
|
+
92 'inputs' in kwargs and 'outputs' in kwargs):
|
|
116
|
+
|
|
117
|
+
93 # Graph network
|
|
118
|
+
|
|
119
|
+
---> 94 self._init_graph_network(*args, **kwargs)
|
|
120
|
+
|
|
121
|
+
95 else:
|
|
122
|
+
|
|
123
|
+
96 # Subclassed network
|
|
124
|
+
|
|
125
|
+
|
|
126
|
+
|
|
127
|
+
~\Anaconda3\lib\site-packages\keras\engine\network.py in _init_graph_network(self, inputs, outputs, name, **kwargs)
|
|
128
|
+
|
|
129
|
+
239 # Keep track of the network's nodes and layers.
|
|
130
|
+
|
|
131
|
+
240 nodes, nodes_by_depth, layers, layers_by_depth = _map_graph_network(
|
|
132
|
+
|
|
133
|
+
--> 241 self.inputs, self.outputs)
|
|
134
|
+
|
|
135
|
+
242 self._network_nodes = nodes
|
|
136
|
+
|
|
137
|
+
243 self._nodes_by_depth = nodes_by_depth
|
|
138
|
+
|
|
139
|
+
|
|
140
|
+
|
|
141
|
+
~\Anaconda3\lib\site-packages\keras\engine\network.py in _map_graph_network(inputs, outputs)
|
|
142
|
+
|
|
143
|
+
1432 layer=layer,
|
|
144
|
+
|
|
145
|
+
1433 node_index=node_index,
|
|
146
|
+
|
|
147
|
+
-> 1434 tensor_index=tensor_index)
|
|
148
|
+
|
|
149
|
+
1435
|
|
150
|
+
|
|
151
|
+
1436 for node in reversed(nodes_in_decreasing_depth):
|
|
152
|
+
|
|
153
|
+
|
|
154
|
+
|
|
155
|
+
~\Anaconda3\lib\site-packages\keras\engine\network.py in build_map(tensor, finished_nodes, nodes_in_progress, layer, node_index, tensor_index)
|
|
156
|
+
|
|
157
|
+
1413
|
|
158
|
+
|
|
159
|
+
1414 # Propagate to all previous tensors connected to this node.
|
|
160
|
+
|
|
161
|
+
-> 1415 for i in range(len(node.inbound_layers)):
|
|
162
|
+
|
|
163
|
+
1416 x = node.input_tensors[i]
|
|
164
|
+
|
|
165
|
+
1417 layer = node.inbound_layers[i]
|
|
166
|
+
|
|
167
|
+
|
|
168
|
+
|
|
169
|
+
TypeError: object of type 'Reshape' has no len()
|
|
170
|
+
|
|
75
171
|
|
|
76
172
|
|
|
77
173
|
エラーメッセージ
|
|
78
174
|
|
|
79
|
-
|
|
175
|
+
|
|
80
|
-
|
|
81
|
-
Model: "sequential_4"
|
|
82
|
-
|
|
83
|
-
_________________________________________________________________
|
|
84
|
-
|
|
85
|
-
Layer (type) Output Shape Param #
|
|
86
|
-
|
|
87
|
-
=================================================================
|
|
88
|
-
|
|
89
|
-
reshape_4 (Reshape) (None, 28, 28, 1) 0
|
|
90
|
-
|
|
91
|
-
_________________________________________________________________
|
|
92
|
-
|
|
93
|
-
conv2d_5 (Conv2D) (None, 28, 28, 16) 416
|
|
94
|
-
|
|
95
|
-
_________________________________________________________________
|
|
96
|
-
|
|
97
|
-
max_pooling2d_4 (MaxPooling2 (None, 14, 14, 16) 0
|
|
98
|
-
|
|
99
|
-
_________________________________________________________________
|
|
100
|
-
|
|
101
|
-
flatten_1 (Flatten) (None, 3136) 0
|
|
102
|
-
|
|
103
|
-
_________________________________________________________________
|
|
104
|
-
|
|
105
|
-
dense_4 (Dense) (None, 2) 6274
|
|
106
|
-
|
|
107
|
-
=================================================================
|
|
108
|
-
|
|
109
|
-
Total params: 6,690
|
|
110
|
-
|
|
111
|
-
Trainable params: 6,690
|
|
112
|
-
|
|
113
|
-
Non-trainable params: 0
|
|
114
|
-
|
|
115
|
-
_________________________________________________________________
|
|
116
|
-
|
|
117
|
-
(5, 5, 1, 16)
|
|
118
|
-
|
|
119
|
-
[[ 0.06888594 0.08682878 0.07998873 0.10239454 0.08470815]
|
|
120
|
-
|
|
121
|
-
[ 0.01193486 0.12805724 -0.00094129 -0.02442928 -0.00218746]
|
|
122
|
-
|
|
123
|
-
[ 0.10281097 0.06729689 0.06930522 0.03168417 0.04593635]
|
|
124
|
-
|
|
125
|
-
[-0.02485886 0.11871461 0.12010761 0.06120927 0.02146379]
|
|
126
|
-
|
|
127
|
-
[ 0.00464662 0.06256669 0.05925997 -0.05820978 -0.02884234]]
|
|
128
176
|
|
|
129
177
|
### 該当のソースコード
|
|
130
178
|
|