python - 无法在 tf.graph (saved_model) 中定位输入张量
问题描述
训练后，我将模型保存为 saved_model 格式（我想将其保存为这种格式而不是 .h5）。加载模型并打印图形时，我找不到输入张量（只有 serving_default_input）。能够做出预测。
keras.applications.VGG16第一次，我使用然后添加了我的模型，keras.Input()但没有任何改变。
这就是我定义模型的方式：
model = keras.applications.VGG16(weights = "imagenet",
include_top = False,
input_shape = (IMG_SIZE[0],IMG_SIZE[1], 3))
for layer in model.layers:
layer.trainable = False
x = model.output
x = Dense(16 , activation="relu")(x)
x = Flatten()(x)
predictions = Dense(1, activation = "sigmoid")(x)
model = Model(inputs = model.input, outputs = predictions)
model.summary() #in the first attempt :
Model: "model_3"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input_3 (InputLayer) [(None, 512, 512, 3)] 0
_________________________________________________________________
block1_conv1 (Conv2D) (None, 512, 512, 64) 1792
_________________________________________________________________
block1_conv2 (Conv2D) (None, 512, 512, 64) 36928
_________________________________________________________________
block1_pool (MaxPooling2D) (None, 256, 256, 64) 0
_________________________________________________________________
block2_conv1 (Conv2D) (None, 256, 256, 128) 73856
_________________________________________________________________
block2_conv2 (Conv2D) (None, 256, 256, 128) 147584
_________________________________________________________________
block2_pool (MaxPooling2D) (None, 128, 128, 128) 0
_________________________________________________________________
block3_conv1 (Conv2D) (None, 128, 128, 256) 295168
_________________________________________________________________
block3_conv2 (Conv2D) (None, 128, 128, 256) 590080
_________________________________________________________________
block3_conv3 (Conv2D) (None, 128, 128, 256) 590080
_________________________________________________________________
block3_pool (MaxPooling2D) (None, 64, 64, 256) 0
_________________________________________________________________
block4_conv1 (Conv2D) (None, 64, 64, 512) 1180160
_________________________________________________________________
block4_conv2 (Conv2D) (None, 64, 64, 512) 2359808
_________________________________________________________________
block4_conv3 (Conv2D) (None, 64, 64, 512) 2359808
_________________________________________________________________
block4_pool (MaxPooling2D) (None, 32, 32, 512) 0
_________________________________________________________________
block5_conv1 (Conv2D) (None, 32, 32, 512) 2359808
_________________________________________________________________
block5_conv2 (Conv2D) (None, 32, 32, 512) 2359808
_________________________________________________________________
block5_conv3 (Conv2D) (None, 32, 32, 512) 2359808
_________________________________________________________________
block5_pool (MaxPooling2D) (None, 16, 16, 512) 0
_________________________________________________________________
dense_5 (Dense) (None, 16, 16, 16) 8208
_________________________________________________________________
flatten_3 (Flatten) (None, 4096) 0
_________________________________________________________________
dense_6 (Dense) (None, 1) 4097
=================================================================
Total params: 14,726,993
Trainable params: 12,305
Non-trainable params: 14,714,688
_________________________________________________________________
None
model.summary() #in the second attempt :
Model: "model_1"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input_5 (InputLayer) [(None, 512, 512, 3)] 0
_________________________________________________________________
vgg16 (Model) (None, 16, 16, 512) 14714688
_________________________________________________________________
dense_3 (Dense) (None, 16, 16, 16) 8208
_________________________________________________________________
flatten_2 (Flatten) (None, 4096) 0
_________________________________________________________________
dense_4 (Dense) (None, 1) 4097
=================================================================
Total params: 14,726,993
Trainable params: 12,305
Non-trainable params: 14,714,688
_________________________________________________________________
None
从 keras 的角度来看。转换为 SavedModel 后，
tf.reset_default_graph()
graph = tf.Graph()
sess = tf.Session(graph=graph)
tf.saved_model.loader.load(sess, [tf.saved_model.SERVING], "SavedModel")
sess.graph.get_operations()
[<tf.Operation 'dense_3_1/kernel' type=VarHandleOp>,
<tf.Operation 'dense_3_1/kernel/Read/ReadVariableOp' type=ReadVariableOp>,
<tf.Operation 'dense_3_1/bias' type=VarHandleOp>,
<tf.Operation 'dense_3_1/bias/Read/ReadVariableOp' type=ReadVariableOp>,
<tf.Operation 'dense_4_1/kernel' type=VarHandleOp>,
<tf.Operation 'dense_4_1/kernel/Read/ReadVariableOp' type=ReadVariableOp>,
<tf.Operation 'dense_4_1/bias' type=VarHandleOp>,
<tf.Operation 'dense_4_1/bias/Read/ReadVariableOp' type=ReadVariableOp>,
<tf.Operation 'block1_conv1/kernel' type=VarHandleOp>,
<tf.Operation 'block1_conv1/kernel/Read/ReadVariableOp' type=ReadVariableOp>,
<tf.Operation 'block1_conv1/bias' type=VarHandleOp>,
<tf.Operation 'block1_conv1/bias/Read/ReadVariableOp' type=ReadVariableOp>,
<tf.Operation 'block1_conv2/kernel' type=VarHandleOp>,
<tf.Operation 'block1_conv2/kernel/Read/ReadVariableOp' type=ReadVariableOp>,
<tf.Operation 'block1_conv2/bias' type=VarHandleOp>,
<tf.Operation 'block1_conv2/bias/Read/ReadVariableOp' type=ReadVariableOp>,
<tf.Operation 'block2_conv1/kernel' type=VarHandleOp>,
<tf.Operation 'block2_conv1/kernel/Read/ReadVariableOp' type=ReadVariableOp>,
<tf.Operation 'block2_conv1/bias' type=VarHandleOp>,
<tf.Operation 'block2_conv1/bias/Read/ReadVariableOp' type=ReadVariableOp>,
<tf.Operation 'block2_conv2/kernel' type=VarHandleOp>,
<tf.Operation 'block2_conv2/kernel/Read/ReadVariableOp' type=ReadVariableOp>,
<tf.Operation 'block2_conv2/bias' type=VarHandleOp>,
<tf.Operation 'block2_conv2/bias/Read/ReadVariableOp' type=ReadVariableOp>,
<tf.Operation 'block3_conv1/kernel' type=VarHandleOp>,
<tf.Operation 'block3_conv1/kernel/Read/ReadVariableOp' type=ReadVariableOp>,
<tf.Operation 'block3_conv1/bias' type=VarHandleOp>,
<tf.Operation 'block3_conv1/bias/Read/ReadVariableOp' type=ReadVariableOp>,
<tf.Operation 'block3_conv2/kernel' type=VarHandleOp>,
<tf.Operation 'block3_conv2/kernel/Read/ReadVariableOp' type=ReadVariableOp>,
<tf.Operation 'block3_conv2/bias' type=VarHandleOp>,
<tf.Operation 'block3_conv2/bias/Read/ReadVariableOp' type=ReadVariableOp>,
<tf.Operation 'block3_conv3/kernel' type=VarHandleOp>,
<tf.Operation 'block3_conv3/kernel/Read/ReadVariableOp' type=ReadVariableOp>,
<tf.Operation 'block3_conv3/bias' type=VarHandleOp>,
<tf.Operation 'block3_conv3/bias/Read/ReadVariableOp' type=ReadVariableOp>,
<tf.Operation 'block4_conv1/kernel' type=VarHandleOp>,
<tf.Operation 'block4_conv1/kernel/Read/ReadVariableOp' type=ReadVariableOp>,
<tf.Operation 'block4_conv1/bias' type=VarHandleOp>,
<tf.Operation 'block4_conv1/bias/Read/ReadVariableOp' type=ReadVariableOp>,
<tf.Operation 'block4_conv2/kernel' type=VarHandleOp>,
<tf.Operation 'block4_conv2/kernel/Read/ReadVariableOp' type=ReadVariableOp>,
<tf.Operation 'block4_conv2/bias' type=VarHandleOp>,
<tf.Operation 'block4_conv2/bias/Read/ReadVariableOp' type=ReadVariableOp>,
<tf.Operation 'block4_conv3/kernel' type=VarHandleOp>,
<tf.Operation 'block4_conv3/kernel/Read/ReadVariableOp' type=ReadVariableOp>,
<tf.Operation 'block4_conv3/bias' type=VarHandleOp>,
<tf.Operation 'block4_conv3/bias/Read/ReadVariableOp' type=ReadVariableOp>,
<tf.Operation 'block5_conv1/kernel' type=VarHandleOp>,
<tf.Operation 'block5_conv1/kernel/Read/ReadVariableOp' type=ReadVariableOp>,
<tf.Operation 'block5_conv1/bias' type=VarHandleOp>,
<tf.Operation 'block5_conv1/bias/Read/ReadVariableOp' type=ReadVariableOp>,
<tf.Operation 'block5_conv2/kernel' type=VarHandleOp>,
<tf.Operation 'block5_conv2/kernel/Read/ReadVariableOp' type=ReadVariableOp>,
<tf.Operation 'block5_conv2/bias' type=VarHandleOp>,
<tf.Operation 'block5_conv2/bias/Read/ReadVariableOp' type=ReadVariableOp>,
<tf.Operation 'block5_conv3/kernel' type=VarHandleOp>,
<tf.Operation 'block5_conv3/kernel/Read/ReadVariableOp' type=ReadVariableOp>,
<tf.Operation 'block5_conv3/bias' type=VarHandleOp>,
<tf.Operation 'block5_conv3/bias/Read/ReadVariableOp' type=ReadVariableOp>,
<tf.Operation 'NoOp' type=NoOp>,
<tf.Operation 'Const' type=Const>,
<tf.Operation 'serving_default_input_5' type=Placeholder>,
<tf.Operation 'StatefulPartitionedCall' type=StatefulPartitionedCall>,
<tf.Operation 'saver_filename' type=Placeholder>,
<tf.Operation 'StatefulPartitionedCall_1' type=StatefulPartitionedCall>,
<tf.Operation 'StatefulPartitionedCall_2' type=StatefulPartitionedCall>]
所以当我试图做出预测时：
in_t = sess.graph.get_tensor_by_name('serving_default_input_5:0')
out = sess.graph.get_tensor_by_name('dense_4_1/bias/Read/ReadVariableOp:0')
...
pred = sess.run([out], feed_dict={ in_t: image}) # image has the right shape
如何将形状 (512,512,3) 的图像传递给我加载的 save_model？
提前TY^^
解决方案
如果in_t = sess.graph.get_tensor_by_name('serving_default_input_5:0')不起作用，您可以尝试使用，因为这是默认情况下Placeholder_0赋予的名称。Input
此外，不确定您是否Input在将其传递给Prediction.
请找到以下代码pre-processing：
IMG_SIZE = 512
img_array = cv2.imread('Image.jpg')
new_array = cv2.resize(img_array, (IMG_SIZE, IMG_SIZE))
new_array = new_array / 255
new_array = new_array.reshape(-1, 512, 512, 3)
下面的代码应该可以工作：
with tf.Session(graph=tf.Graph()) as sess:
tf.saved_model.loader.load(
sess,
[tf.saved_model.tag_constants.SERVING],
"Saved_Model"
)
prediction = sess.run(
[out],
feed_dict={'Placeholder:0': new_array})
print(prediction)
如果您仍然遇到任何错误，请分享完整的错误跟踪，我们将很乐意为您提供帮助。
希望这可以帮助。快乐学习！
推荐阅读
- html - 一个 div 没有出现，它不可见
- amazon-web-services - Jenkins ECR - 无法运行程序“docker”
- python - 带有 7 天周实际开始/结束偏移量的 Python 日期范围
- python - 如何使用 Python Selenium 在没有 ID 或名称的页面上单击“接受 Cookie”
- typeerror - 3 p5 草图尚未加载：TypeError {}
- java - 多态性。复制数组的问题
- wpf - Squirrel.Windows 增量更新不起作用
- typescript - 如何描述从 .ts 文件导出为函数的 webpack 配置类型？
- python - 尝试使用列表实现斐波那契数列的迭代解决方案时出现问题
- json.net - 无法创建 ITask 类型的实例。类型是接口或抽象类，不能实例化