Is it possible to print the output of a specific layer of a tensorflow model? - tensorflow

Question

I constructed this dataset for binary classification that contains digit 0 vs. digit 6.
import tensorflow as tf
from sklearn import datasets
from sklearn.model_selection import train_test_split
import numpy as np
mnist = datasets.load_digits()
# generate the indices
idx_digit = np.argwhere((mnist.target == 0) | (mnist.target == 6)).flatten()
X_train, X_test, y_train, y_test = train_test_split(
mnist.data[idx_digit].reshape((-1,8,8,1)),
mnist.target[idx_digit], test_size=0.33, random_state=42)
y_train[y_train==6]=1
y_test[y_test==6]=1
I built a convolutional neural network with keras.
model = tf.keras.Sequential([
tf.keras.layers.Conv2D(filters=1, kernel_size=(3, 3), activation='relu'),
tf.keras.layers.AveragePooling2D(),
tf.keras.layers.Flatten(),
tf.keras.layers.Dense(1, activation=tf.nn.sigmoid)
])
I compiled and trained the model and it works well.
model.compile(loss=tf.keras.losses.binary_crossentropy,
optimizer=tf.keras.optimizers.Adam(),
metrics=['accuracy'])
model.fit(X_train, y_train, epochs=1, verbose=1)
I'd just like to know if it is possible to print the output of a specific layer, e.g.
model.layers[0].output
In other words, how do I get the output of the convolutional layer for a given input x?

Answer 1

In order to get the output of the intermediate layer after training a model, you can do
Along with the last layer output, if we want to receive the convolutional layer's output (of yours).
feature_extractor = tf.keras.Model(
inputs=model.inputs,
outputs=[
model.output, # < last layer output
model.layers[0].output # < your convolution layer output
]
)
x = tf.ones((1, 8, 8, 1))
y, conv_y = feature_extractor(x)
y.shape, conv_y.shape
(TensorShape([1, 1]), TensorShape([1, 6, 6, 1]))
Also, if we want to get all layers output, then we can do
feature_extractor = tf.keras.Model(
inputs=model.inputs,
outputs=[layer.output for layer in model.layers],
)
features = feature_extractor(x); print(len(features))
4
for i in range(len(features)):
print(features[i].shape)
(1, 6, 6, 1) < first layer output / conv layer
(1, 3, 3, 1) < second layer output
(1, 9) < 3rd and
(1, 1) < 4th (last layer)