吳恩達深度學習筆記 course4 week1 做業2
Residual Networks
Welcome to the second assignment of this week! You will learn how to build very deep convolutional networks, using Residual Networks (ResNets). In theory, very deep networks can represent very complex functions; but in practice, they are hard to train. Residual Networks, introduced by He et al., allow you to train much deeper networks than were previously practically feasible.css
In this assignment, you will:html
- Implement the basic building blocks of ResNets.
- Put together these building blocks to implement and train a state-of-the-art neural network for image classification.
This assignment will be done in Keras.python
Before jumping into the problem, let's run the cell below to load the required packages.git
import numpy as np
from keras import layers
from keras.layers import Input, Add, Dense, Activation, ZeroPadding2D, BatchNormalization, Flatten, Conv2D, AveragePooling2D, MaxPooling2D, GlobalMaxPooling2D
from keras.models import Model, load_model
from keras.preprocessing import image
from keras.utils import layer_utils
from keras.utils.data_utils import get_file
from keras.applications.imagenet_utils import preprocess_input
import pydot
from IPython.display import SVG
from keras.utils.vis_utils import model_to_dot
from keras.utils import plot_model
from resnets_utils import *
from keras.initializers import glorot_uniform
import scipy.misc
from matplotlib.pyplot import imshow
%matplotlib inline
import keras.backend as K
K.set_image_data_format('channels_last')
K.set_learning_phase(1)
1 - The problem of very deep neural networks
Last week, you built your first convolutional neural network. In recent years, neural networks have become deeper, with state-of-the-art networks going from just a few layers (e.g., AlexNet) to over a hundred layers.github
The main benefit of a very deep network is that it can represent very complex functions. It can also learn features at many different levels of abstraction, from edges (at the lower layers) to very complex features (at the deeper layers). However, using a deeper network doesn't always help. A huge barrier to training them is vanishing gradients: very deep networks often have a gradient signal that goes to zero quickly, thus making gradient descent unbearably slow. More specifically, during gradient descent, as you backprop from the final layer back to the first layer, you are multiplying by the weight matrix on each step, and thus the gradient can decrease exponentially quickly to zero (or, in rare cases, grow exponentially quickly and "explode" to take very large values).算法
During training, you might therefore see the magnitude (or norm) of the gradient for the earlier layers descrease to zero very rapidly as training proceeds:api
The speed of learning decreases very rapidly for the early layers as the network trains
You are now going to solve this problem by building a Residual Network!markdown
2 - Building a Residual Network
In ResNets, a "shortcut" or a "skip connection" allows the gradient to be directly backpropagated to earlier layers:網絡
The image on the left shows the "main path" through the network. The image on the right adds a shortcut to the main path. By stacking these ResNet blocks on top of each other, you can form a very deep network.app
We also saw in lecture that having ResNet blocks with the shortcut also makes it very easy for one of the blocks to learn an identity function. This means that you can stack on additional ResNet blocks with little risk of harming training set performance. (There is also some evidence that the ease of learning an identity function--even more than skip connections helping with vanishing gradients--accounts for ResNets' remarkable performance.)
Two main types of blocks are used in a ResNet, depending mainly on whether the input/output dimensions are same or different. You are going to implement both of them.
2.1 - The identity block
The identity block is the standard block used in ResNets, and corresponds to the case where the input activation (say a[l]a[l]) has the same dimension as the output activation (say a[l+2]a[l+2]). To flesh out the different steps of what happens in a ResNet's identity block, here is an alternative diagram showing the individual steps:
The upper path is the "shortcut path." The lower path is the "main path." In this diagram, we have also made explicit the CONV2D and ReLU steps in each layer. To speed up training we have also added a BatchNorm step. Don't worry about this being complicated to implement--you'll see that BatchNorm is just one line of code in Keras!
In this exercise, you'll actually implement a slightly more powerful version of this identity block, in which the skip connection "skips over" 3 hidden layers rather than 2 layers. It looks like this:
Here're the individual steps.
First component of main path:
- The first CONV2D has F1F1 filters of shape (1,1) and a stride of (1,1). Its padding is "valid" and its name should be
conv_name_base + '2a'. Use 0 as the seed for the random initialization. - The first BatchNorm is normalizing the channels axis. Its name should be
bn_name_base + '2a'. - Then apply the ReLU activation function. This has no name and no hyperparameters.
Second component of main path:
- The second CONV2D has F2F2 filters of shape (f,f)(f,f) and a stride of (1,1). Its padding is "same" and its name should be
conv_name_base + '2b'. Use 0 as the seed for the random initialization. - The second BatchNorm is normalizing the channels axis. Its name should be
bn_name_base + '2b'. - Then apply the ReLU activation function. This has no name and no hyperparameters.
Third component of main path:
- The third CONV2D has F3F3 filters of shape (1,1) and a stride of (1,1). Its padding is "valid" and its name should be
conv_name_base + '2c'. Use 0 as the seed for the random initialization. - The third BatchNorm is normalizing the channels axis. Its name should be
bn_name_base + '2c'. Note that there is no ReLU activation function in this component.
Final step:
- The shortcut and the input are added together.
- Then apply the ReLU activation function. This has no name and no hyperparameters.
Exercise: Implement the ResNet identity block. We have implemented the first component of the main path. Please read over this carefully to make sure you understand what it is doing. You should implement the rest.
- To implement the Conv2D step: See reference
- To implement BatchNorm: See reference (axis: Integer, the axis that should be normalized (typically the channels axis))
- For the activation, use:
Activation('relu')(X) - To add the value passed forward by the shortcut: See reference
# GRADED FUNCTION: identity_block
def identity_block(X, f, filters, stage, block):
"""
Implementation of the identity block as defined in Figure 3
Arguments:
X -- input tensor of shape (m, n_H_prev, n_W_prev, n_C_prev)
f -- integer, specifying the shape of the middle CONV's window for the main path
filters -- python list of integers, defining the number of filters in the CONV layers of the main path
stage -- integer, used to name the layers, depending on their position in the network
block -- string/character, used to name the layers, depending on their position in the network
Returns:
X -- output of the identity block, tensor of shape (n_H, n_W, n_C)
"""
# defining name basis
conv_name_base = 'res' + str(stage) + block + '_branch'
bn_name_base = 'bn' + str(stage) + block + '_branch'
# Retrieve Filters
F1, F2, F3 = filters
# Save the input value. You'll need this later to add back to the main path.
X_shortcut = X
# First component of main path
X = Conv2D(filters = F1, kernel_size = (1, 1), strides = (1,1), padding = 'valid', name = conv_name_base + '2a', kernel_initializer = glorot_uniform(seed=0))(X)
X = BatchNormalization(axis = 3, name = bn_name_base + '2a')(X)
X = Activation('relu')(X)
### START CODE HERE ###
# Second component of main path (≈3 lines)
X = Conv2D(filters = F2, kernel_size = (f, f), strides = (1,1), padding = 'same', name = conv_name_base + '2b', kernel_initializer = glorot_uniform(seed=0))(X)
X =BatchNormalization(axis = 3, name = bn_name_base + '2b')(X)
X = Activation('relu')(X)
# Third component of main path (≈2 lines)
X = Conv2D(filters = F3, kernel_size = (1, 1), strides = (1,1), padding = 'same', name = conv_name_base + '2c', kernel_initializer = glorot_uniform(seed=0))(X)
X = BatchNormalization(axis = 3, name = bn_name_base + '2c')(X)
# Final step: Add shortcut value to main path, and pass it through a RELU activation (≈2 lines)
X = Add()([X,X_shortcut])
X = Activation('relu')(X)
### END CODE HERE ###
return X
tf.reset_default_graph()
with tf.Session() as test:
np.random.seed(1)
A_prev = tf.placeholder("float", [3, 4, 4, 6])
X = np.random.randn(3, 4, 4, 6)
A = identity_block(A_prev, f = 2, filters = [2, 4, 6], stage = 1, block = 'a')
test.run(tf.global_variables_initializer())
out = test.run([A], feed_dict={A_prev: X, K.learning_phase(): 0})
print("out = " + str(out[0][1][1][0]))
Expected Output:
| out | [ 0.94822985 0. 1.16101444 2.747859 0. 1.36677003] |
2.2 - The convolutional block
You've implemented the ResNet identity block. Next, the ResNet "convolutional block" is the other type of block. You can use this type of block when the input and output dimensions don't match up. The difference with the identity block is that there is a CONV2D layer in the shortcut path:
The CONV2D layer in the shortcut path is used to resize the input xx to a different dimension, so that the dimensions match up in the final addition needed to add the shortcut value back to the main path. (This plays a similar role as the matrix WsWs discussed in lecture.) For example, to reduce the activation dimensions's height and width by a factor of 2, you can use a 1x1 convolution with a stride of 2. The CONV2D layer on the shortcut path does not use any non-linear activation function. Its main role is to just apply a (learned) linear function that reduces the dimension of the input, so that the dimensions match up for the later addition step.
The details of the convolutional block are as follows.
First component of main path:
- The first CONV2D has F1F1 filters of shape (1,1) and a stride of (s,s). Its padding is "valid" and its name should be
conv_name_base + '2a'. - The first BatchNorm is normalizing the channels axis. Its name should be
bn_name_base + '2a'. - Then apply the ReLU activation function. This has no name and no hyperparameters.
Second component of main path:
- The second CONV2D has F2F2 filters of (f,f) and a stride of (1,1). Its padding is "same" and it's name should be
conv_name_base + '2b'. - The second BatchNorm is normalizing the channels axis. Its name should be
bn_name_base + '2b'. - Then apply the ReLU activation function. This has no name and no hyperparameters.
Third component of main path:
- The third CONV2D has F3F3 filters of (1,1) and a stride of (1,1). Its padding is "valid" and it's name should be
conv_name_base + '2c'. - The third BatchNorm is normalizing the channels axis. Its name should be
bn_name_base + '2c'. Note that there is no ReLU activation function in this component.
Shortcut path:
- The CONV2D has F3F3 filters of shape (1,1) and a stride of (s,s). Its padding is "valid" and its name should be
conv_name_base + '1'. - The BatchNorm is normalizing the channels axis. Its name should be
bn_name_base + '1'.
Final step:
- The shortcut and the main path values are added together.
- Then apply the ReLU activation function. This has no name and no hyperparameters.
Exercise: Implement the convoluti onal block. We have implemented the first component of the main path; you should implement the rest. As before, always use 0 as the seed for the random initialization, to ensure consistency with our grader.
- Conv Hint
- BatchNorm Hint (axis: Integer, the axis that should be normalized (typically the features axis))
- For the activation, use:
Activation('relu')(X) - Addition Hint
# GRADED FUNCTION: convolutional_block
def convolutional_block(X, f, filters, stage, block, s = 2):
"""
Implementation of the convolutional block as defined in Figure 4
Arguments:
X -- input tensor of shape (m, n_H_prev, n_W_prev, n_C_prev)
f -- integer, specifying the shape of the middle CONV's window for the main path
filters -- python list of integers, defining the number of filters in the CONV layers of the main path
stage -- integer, used to name the layers, depending on their position in the network
block -- string/character, used to name the layers, depending on their position in the network
s -- Integer, specifying the stride to be used
Returns:
X -- output of the convolutional block, tensor of shape (n_H, n_W, n_C)
"""
# defining name basis
conv_name_base = 'res' + str(stage) + block + '_branch'
bn_name_base = 'bn' + str(stage) + block + '_branch'
# Retrieve Filters
F1, F2, F3 = filters
# Save the input value
X_shortcut = X
##### MAIN PATH #####
# First component of main path
X = Conv2D(F1, kernel_size = (1, 1), strides = (s,s), name = conv_name_base + '2a',padding = 'valid', kernel_initializer = glorot_uniform(seed=0))(X)
X = BatchNormalization(axis = 3, name = bn_name_base + '2a')(X)
X = Activation('relu')(X)
### START CODE HERE ###
# Second component of main path (≈3 lines)
X = Conv2D(F2, kernel_size = (f, f), strides = (1,1), name = conv_name_base + '2b',padding = 'same', kernel_initializer = glorot_uniform(seed=0))(X)
X = BatchNormalization(axis = 3, name = bn_name_base + '2b')(X)
X = Activation('relu')(X)
# Third component of main path (≈2 lines)
X = Conv2D(F3, kernel_size = (1, 1), strides = (1,1), name = conv_name_base + '2c',padding = 'valid', kernel_initializer = glorot_uniform(seed=0))(X)
X = BatchNormalization(axis = 3, name = bn_name_base + '2c')(X)
##### SHORTCUT PATH #### (≈2 lines)
X_shortcut = Conv2D(F3, kernel_size = (1, 1), strides = (s,s), name = conv_name_base + '1',padding = 'valid', kernel_initializer = glorot_uniform(seed=0))(X_shortcut)
X_shortcut = BatchNormalization(axis = 3, name = bn_name_base + '1')(X_shortcut)
# Final step: Add shortcut value to main path, and pass it through a RELU activation (≈2 lines)
X = Add()([X,X_shortcut])
X = Activation('relu')(X)
### END CODE HERE ###
return X
tf.reset_default_graph()
with tf.Session() as test:
np.random.seed(1)
A_prev = tf.placeholder("float", [3, 4, 4, 6])
X = np.random.randn(3, 4, 4, 6)
A = convolutional_block(A_prev, f = 2, filters = [2, 4, 6], stage = 1, block = 'a')
test.run(tf.global_variables_initializer())
out = test.run([A], feed_dict={A_prev: X, K.learning_phase(): 0})
print("out = " + str(out[0][1][1][0]))
Expected Output:
| out | [ 0.09018463 1.23489773 0.46822017 0.0367176 0. 0.65516603] |
3 - Building your first ResNet model (50 layers)
You now have the necessary blocks to build a very deep ResNet. The following figure describes in detail the architecture of this neural network. "ID BLOCK" in the diagram stands for "Identity block," and "ID BLOCK x3" means you should stack 3 identity blocks together.
The details of this ResNet-50 model are:
- Zero-padding pads the input with a pad of (3,3)
- Stage 1:
- The 2D Convolution has 64 filters of shape (7,7) and uses a stride of (2,2). Its name is "conv1".
- BatchNorm is applied to the channels axis of the input.
- MaxPooling uses a (3,3) window and a (2,2) stride.
- Stage 2:
- The convolutional block uses three set of filters of size [64,64,256], "f" is 3, "s" is 1 and the block is "a".
- The 2 identity blocks use three set of filters of size [64,64,256], "f" is 3 and the blocks are "b" and "c".
- Stage 3:
- The convolutional block uses three set of filters of size [128,128,512], "f" is 3, "s" is 2 and the block is "a".
- The 3 identity blocks use three set of filters of size [128,128,512], "f" is 3 and the blocks are "b", "c" and "d".
- Stage 4:
- The convolutional block uses three set of filters of size [256, 256, 1024], "f" is 3, "s" is 2 and the block is "a".
- The 5 identity blocks use three set of filters of size [256, 256, 1024], "f" is 3 and the blocks are "b", "c", "d", "e" and "f".
- Stage 5:
- The convolutional block uses three set of filters of size [512, 512, 2048], "f" is 3, "s" is 2 and the block is "a".
- The 2 identity blocks use three set of filters of size [512, 512, 2048], "f" is 3 and the blocks are "b" and "c".
- The 2D Average Pooling uses a window of shape (2,2) and its name is "avg_pool".
- The flatten doesn't have any hyperparameters or name.
- The Fully Connected (Dense) layer reduces its input to the number of classes using a softmax activation. Its name should be
'fc' + str(classes).
Exercise: Implement the ResNet with 50 layers described in the figure above. We have implemented Stages 1 and 2. Please implement the rest. (The syntax for implementing Stages 3-5 should be quite similar to that of Stage 2.) Make sure you follow the naming convention in the text above.
You'll need to use this function:
- Average pooling see reference
Here're some other functions we used in the code below:
- Conv2D: See reference
- BatchNorm: See reference (axis: Integer, the axis that should be normalized (typically the features axis))
- Zero padding: See reference
- Max pooling: See reference
- Fully conected layer: See reference
- Addition: See reference
# GRADED FUNCTION: ResNet50
def ResNet50(input_shape = (64, 64, 3), classes = 6):
"""
Implementation of the popular ResNet50 the following architecture:
CONV2D -> BATCHNORM -> RELU -> MAXPOOL -> CONVBLOCK -> IDBLOCK*2 -> CONVBLOCK -> IDBLOCK*3
-> CONVBLOCK -> IDBLOCK*5 -> CONVBLOCK -> IDBLOCK*2 -> AVGPOOL -> TOPLAYER
Arguments:
input_shape -- shape of the images of the dataset
classes -- integer, number of classes
Returns:
model -- a Model() instance in Keras
"""
# Define the input as a tensor with shape input_shape
X_input = Input(input_shape)
# Zero-Padding
X = ZeroPadding2D((3, 3))(X_input)
# Stage 1
X = Conv2D(64, (7, 7), strides = (2, 2), name = 'conv1', kernel_initializer = glorot_uniform(seed=0))(X)
X = BatchNormalization(axis = 3, name = 'bn_conv1')(X)
X = Activation('relu')(X)
X = MaxPooling2D((3, 3), strides=(2, 2))(X)
# Stage 2
X = convolutional_block(X, f = 3, filters = [64, 64, 256], stage = 2, block='a', s = 1)
X = identity_block(X, 3, [64, 64, 256], stage=2, block='b')
X = identity_block(X, 3, [64, 64, 256], stage=2, block='c')
### START CODE HERE ###
# Stage 3 (≈4 lines)
X = convolutional_block(X, f = 3, filters = [128, 128, 512], stage = 3, block='a', s = 2)
X = identity_block(X, 3, [128, 128, 512], stage=3, block='b')
X = identity_block(X, 3, [128, 128, 512], stage=3, block='c')
X = identity_block(X, 3, [128, 128, 512], stage=3, block='d')
# Stage 4 (≈6 lines)
X = convolutional_block(X, f = 3, filters = [256, 256, 1024], stage = 4, block='a', s = 2)
X = identity_block(X, 3, [256, 256, 1024], stage=4, block='b')
X = identity_block(X, 3, [256, 256, 1024], stage=4, block='c')
X = identity_block(X, 3, [256, 256, 1024], stage=4, block='d')
X = identity_block(X, 3, [256, 256, 1024], stage=4, block='e')
X = identity_block(X, 3, [256, 256, 1024], stage=4, block='f')
# Stage 5 (≈3 lines)
X = convolutional_block(X, f = 3, filters = [512, 512, 2048], stage = 5, block='a', s = 2)
X = identity_block(X, 3, [512, 512, 2048], stage=5, block='b')
X = identity_block(X, 3, [512, 512, 2048], stage=5, block='c')
# AVGPOOL (≈1 line). Use "X = AveragePooling2D(...)(X)"
X = AveragePooling2D(pool_size=(2, 2), padding='same')(X)
### END CODE HERE ###
# output layer
X = Flatten()(X)
X = Dense(classes, activation='softmax', name='fc' + str(classes), kernel_initializer = glorot_uniform(seed=0))(X)
# Create model
model = Model(inputs = X_input, outputs = X, name='ResNet50')
return model
Run the following code to build the model's graph. If your implementation is not correct you will know it by checking your accuracy when running model.fit(...)below.
model = ResNet50(input_shape = (64, 64, 3), classes = 6)
As seen in the Keras Tutorial Notebook, prior training a model, you need to configure the learning process by compiling the model.
model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])
The model is now ready to be trained. The only thing you need is a dataset.
Let's load the SIGNS Dataset.
X_train_orig, Y_train_orig, X_test_orig, Y_test_orig, classes = load_dataset()
# Normalize image vectors
X_train = X_train_orig/255.
X_test = X_test_orig/255.
# Convert training and test labels to one hot matrices
Y_train = convert_to_one_hot(Y_train_orig, 6).T
Y_test = convert_to_one_hot(Y_test_orig, 6).T
print ("number of training examples = " + str(X_train.shape[0]))
print ("number of test examples = " + str(X_test.shape[0]))
print ("X_train shape: " + str(X_train.shape))
print ("Y_train shape: " + str(Y_train.shape))
print ("X_test shape: " + str(X_test.shape))
print ("Y_test shape: " + str(Y_test.shape))
Run the following cell to train your model on 2 epochs with a batch size of 32. On a CPU it should take you around 5min per epoch.
model.fit(X_train, Y_train, epochs = 40, batch_size = 32)
Expected Output:
| Epoch 1/2 | loss: between 1 and 5, acc: between 0.2 and 0.5, although your results can be different from ours. |
| Epoch 2/2 | loss: between 1 and 5, acc: between 0.2 and 0.5, you should see your loss decreasing and the accuracy increasing. |
Let's see how this model (trained on only two epochs) performs on the test set.
preds = model.evaluate(X_test, Y_test)
print ("Loss = " + str(preds[0]))
print ("Test Accuracy = " + str(preds[1]))
Expected Output:
| Test Accuracy | between 0.16 and 0.25 |
For the purpose of this assignment, we've asked you to train the model only for two epochs. You can see that it achieves poor performances. Please go ahead and submit your assignment; to check correctness, the online grader will run your code only for a small number of epochs as well.
After you have finished this official (graded) part of this assignment, you can also optionally train the ResNet for more iterations, if you want. We get a lot better performance when we train for ~20 epochs, but this will take more than an hour when training on a CPU.
Using a GPU, we've trained our own ResNet50 model's weights on the SIGNS dataset. You can load and run our trained model on the test set in the cells below. It may take ≈1min to load the model.
model = load_model('ResNet50.h5')
preds = model.evaluate(X_test, Y_test)
print ("Loss = " + str(preds[0]))
print ("Test Accuracy = " + str(preds[1]))
ResNet50 is a powerful model for image classification when it is trained for an adequate number of iterations. We hope you can use what you've learnt and apply it to your own classification problem to perform state-of-the-art accuracy.
Congratulations on finishing this assignment! You've now implemented a state-of-the-art image classification system!
4 - Test on your own image (Optional/Ungraded)
If you wish, you can also take a picture of your own hand and see the output of the model. To do this:
1. Click on "File" in the upper bar of this notebook, then click "Open" to go on your Coursera Hub. 2. Add your image to this Jupyter Notebook's directory, in the "images" folder 3. Write your image's name in the following code 4. Run the code and check if the algorithm is right!
img_path = 'images/my_image.jpg'
img = image.load_img(img_path, target_size=(64, 64))
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
x = preprocess_input(x)
print('Input image shape:', x.shape)
my_image = scipy.misc.imread(img_path)
imshow(my_image)
print("class prediction vector [p(0), p(1), p(2), p(3), p(4), p(5)] = ")
print(model.predict(x))
You can also print a summary of your model by running the following code.
model.summary()
Finally, run the code below to visualize your ResNet50. You can also download a .png picture of your model by going to "File -> Open...-> model.png".
plot_model(model, to_file='model.png')
SVG(model_to_dot(model).create(prog='dot', format='svg'))
What you should remember:
- Very deep "plain" networks don't work in practice because they are hard to train due to vanishing gradients.
- The skip-connections help to address the Vanishing Gradient problem. They also make it easy for a ResNet block to learn an identity function.
- There are two main type of blocks: The identity block and the convolutional block.
- Very deep Residual Networks are built by stacking these blocks together.
References
This notebook presents the ResNet algorithm due to He et al. (2015). The implementation here also took significant inspiration and follows the structure given in the github repository of Francois Chollet:
- Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun - Deep Residual Learning for Image Recognition (2015)
- Francois Chollet's github repository: https://github.com/fchollet/deep-learning-models/blob/master/resnet50.py
-------------------------------------------------------------中文版---------------------------------------------------------------------------------------------
中文版摘自:https://blog.csdn.net/u013733326
殘差網絡的搭建
這裏咱們將學習怎樣使用殘差網絡構建一個很是深的卷積網絡。理論上越深的網絡越可以實現越複雜的功能，可是在實際上卻很是難以訓練。殘差網絡就是爲了解決深網絡的難以訓練的問題的。
在本文章中，咱們將：
- 實現基本的殘差塊。
- 將這些殘差塊放在一塊兒，實現並訓練用於圖像分類的神經網絡。
本次實驗將使用Keras框架
在解決問題以前，咱們先來導入庫函數：
import numpy as np import tensorflow as tf from keras import layers from keras.layers import Input, Add, Dense, Activation, ZeroPadding2D, BatchNormalization, Flatten, Conv2D, AveragePooling2D, MaxPooling2D, GlobalMaxPooling2D from keras.models import Model, load_model from keras.preprocessing import image from keras.utils import layer_utils from keras.utils.data_utils import get_file from keras.applications.imagenet_utils import preprocess_input from keras.utils.vis_utils import model_to_dot from keras.utils import plot_model from keras.initializers import glorot_uniform import pydot from IPython.display import SVG import scipy.misc from matplotlib.pyplot import imshow import keras.backend as K K.set_image_data_format('channels_last') K.set_learning_phase(1) import resnets_utils
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10
- 11
- 12
- 13
- 14
- 15
- 16
- 17
- 18
- 19
- 20
- 21
- 22
- 23
2.1 - 深層網絡的麻煩
上週，咱們構建了第一個卷積神經網絡。最近幾年，卷積神經網絡變得愈來愈深，從只從幾層（例如AlexNet）到超過一百層。
使用深層網絡最大的好處就是它可以完成很複雜的功能，它可以從邊緣（淺層）到很是複雜的特徵（深層）中不一樣的抽象層次的特徵中學習。然而，使用比較深的網絡一般沒有什麼好處，一個特別大的麻煩就在於訓練的時候會產生梯度消失，很是深的網絡一般會有一個梯度信號，該信號會迅速的消退，從而使得梯度降低變得很是緩慢。更具體的說，在梯度降低的過程當中，當你從最後一層回到第一層的時候，你在每一個步驟上乘以權重矩陣，所以梯度值能夠迅速的指數式地減小到0（在極少數的狀況下會迅速增加，形成梯度爆炸）。
在訓練的過程當中，你可能會看到開始幾層的梯度的大小（或範數）迅速降低到0，以下圖：
在前幾層中隨着迭代次數的增長，學習的速度會降低的很是快。
爲了解決這個問題，咱們將構建殘差網絡。
2.2 - 構建一個殘差網絡
在殘差網絡中，一個「捷徑（shortcut）」或者說「跳躍鏈接（skip connection）」容許梯度直接反向傳播到更淺的層，以下圖：
圖像左邊是神經網絡的主路，圖像右邊是添加了一條捷徑的主路，經過這些殘差塊堆疊在一塊兒，能夠造成一個很是深的網絡。
咱們在視頻中能夠看到使用捷徑的方式使得每個殘差塊可以很容易學習到恆等式功能，這意味着咱們能夠添加不少的殘差塊而不會損害訓練集的表現。
殘差塊有兩種類型，主要取決於輸入輸出的維度是否相同，下面咱們來看看吧~
2.2.1 - 恆等塊（Identity block）
恆等塊是殘差網絡使用的的標準塊，對應於輸入的激活值（好比a[l]a[l]）與輸出激活值（好比a[l+1]a[l+1]）具備相同的維度。爲了具象化殘差塊的不一樣步驟，咱們來看看下面的圖吧~
上圖中，上面的曲線路徑是「捷徑」，下面的直線路徑是主路徑。在上圖中，咱們依舊把CONV2D 與 ReLU包含到了每一個步驟中，爲了提高訓練的速度，咱們在每一步也把數據進行了歸一化（BatchNorm），不要懼怕這些東西，由於Keras框架已經實現了這些東西，調用BatchNorm只須要一行代碼。
在實踐中，咱們要作一個更強大的版本：跳躍鏈接會跳過3個隱藏層而不是兩個，就像下圖：
每一個步驟以下：
-
主路徑的第一部分：
-
第一個CONV2D有F1F1個過濾器，其大小爲（11，11），步長爲（1，1），使用填充方式爲「valid」，命名規則爲
conv_name_base + '2a'，使用00做爲隨機種子爲其初始化。 -
第一個BatchNorm是通道的軸歸一化，其命名規則爲
bn_name_base + '2a'。 -
接着使用ReLU激活函數，它沒有命名也沒有超參數。
-
-
主路徑的第二部分：
-
第二個CONV2D有F2F2個過濾器，其大小爲（ff，ff），步長爲（1，1），使用填充方式爲「same」，命名規則爲
conv_name_base + '2b'，使用00做爲隨機種子爲其初始化。 -
第二個BatchNorm是通道的軸歸一化，其命名規則爲
bn_name_base + '2b'。 -
接着使用ReLU激活函數，它沒有命名也沒有超參數。
-
-
主路徑的第三部分：
-
第三個CONV2D有F3F3個過濾器，其大小爲（11，11），步長爲（1，1），使用填充方式爲「valid」，命名規則爲
conv_name_base + '2c'，使用00做爲隨機種子爲其初始化。 -
第三個BatchNorm是通道的軸歸一化，其命名規則爲
bn_name_base + '2c'。 -
注意這裏沒有ReLU函數
-
-
最後一步：
-
將捷徑與輸入加在一塊兒
-
使用ReLU激活函數，它沒有命名也沒有超參數。
-
接下來咱們就要實現殘差網絡的恆等塊了，請務必查看下面的中文手冊：
def identity_block(X, f, filters, stage, block): """ 實現圖3的恆等塊 參數： X - 輸入的tensor類型的數據，維度爲( m, n_H_prev, n_W_prev, n_H_prev ) f - 整數，指定主路徑中間的CONV窗口的維度 filters - 整數列表，定義了主路徑每層的卷積層的過濾器數量 stage - 整數，根據每層的位置來命名每一層，與block參數一塊兒使用。 block - 字符串，據每層的位置來命名每一層，與stage參數一塊兒使用。 返回： X - 恆等塊的輸出，tensor類型，維度爲(n_H, n_W, n_C) """ #定義命名規則 conv_name_base = "res" + str(stage) + block + "_branch" bn_name_base = "bn" + str(stage) + block + "_branch" #獲取過濾器 F1, F2, F3 = filters #保存輸入數據，將會用於爲主路徑添加捷徑 X_shortcut = X #主路徑的第一部分 ##卷積層 X = Conv2D(filters=F1, kernel_size=(1,1), strides=(1,1) ,padding="valid", name=conv_name_base+"2a", kernel_initializer=glorot_uniform(seed=0))(X) ##歸一化 X = BatchNormalization(axis=3,name=bn_name_base+"2a")(X) ##使用ReLU激活函數 X = Activation("relu")(X) #主路徑的第二部分 ##卷積層 X = Conv2D(filters=F2, kernel_size=(f,f),strides=(1,1), padding="same", name=conv_name_base+"2b", kernel_initializer=glorot_uniform(seed=0))(X) ##歸一化 X = BatchNormalization(axis=3,name=bn_name_base+"2b")(X) ##使用ReLU激活函數 X = Activation("relu")(X) #主路徑的第三部分 ##卷積層 X = Conv2D(filters=F3, kernel_size=(1,1), strides=(1,1), padding="valid", name=conv_name_base+"2c", kernel_initializer=glorot_uniform(seed=0))(X) ##歸一化 X = BatchNormalization(axis=3,name=bn_name_base+"2c")(X) ##沒有ReLU激活函數 #最後一步： ##將捷徑與輸入加在一塊兒 X = Add()([X,X_shortcut]) ##使用ReLU激活函數 X = Activation("relu")(X) return X
- 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
咱們來測試一下：
tf.reset_default_graph()
with tf.Session() as test: np.random.seed(1) A_prev = tf.placeholder("float",[3,4,4,6]) X = np.random.randn(3,4,4,6) A = identity_block(A_prev,f=2,filters=[2,4,6],stage=1,block="a") test.run(tf.global_variables_initializer()) out = test.run([A],feed_dict={A_prev:X,K.learning_phase():0}) print("out = " + str(out[0][1][1][0])) test.close()
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10
- 11
- 12
測試結果：
out = [ 0.19716813 0. 1.35612273 2.17130733 0. 1.33249867]
- 1
2.2.2 - 卷積塊
咱們已經實現了殘差網絡的恆等塊，如今，殘差網絡的卷積塊是另外一種類型的殘差塊，它適用於輸入輸出的維度不一致的狀況，它不一樣於上面的恆等塊，與之區別在於，捷徑中有一個CONV2D層，以下圖：
捷徑中的卷積層將把輸入xx卷積爲不一樣的維度，所以在主路徑最後那裏須要適配捷徑中的維度。好比：把激活值中的寬高減小2倍，咱們可使用1x1的卷積，步伐爲2。捷徑上的卷積層不使用任何非線性激活函數，它的主要做用是僅僅應用（學習後的）線性函數來減小輸入的維度，以便在後面的加法步驟中的維度相匹配。
具體步驟以下：
-
主路徑第一部分：
-
第一個卷積層有F1F1個過濾器，其維度爲（11，11），步伐爲（ss，ss），使用「valid」的填充方式，命名規則爲
conv_name_base + '2a' -
第一個規範層是通道的軸歸一化，其命名規則爲
bn_name_base + '2a' -
使用ReLU激活函數，它沒有命名規則也沒有超參數。
-
-
主路徑第二部分：
-
第二個卷積層有F2F2個過濾器，其維度爲（ff，ff），步伐爲（11，11），使用「same」的填充方式，命名規則爲
conv_name_base + '2b' -
第二個規範層是通道的軸歸一化，其命名規則爲
bn_name_base + '2b' -
使用ReLU激活函數，它沒有命名規則也沒有超參數。
-
-
主路徑第三部分：
-
第三個卷積層有F3F3個過濾器，其維度爲（11，11），步伐爲（ss，ss），使用「valid」的填充方式，命名規則爲
conv_name_base + '2c' -
第三個規範層是通道的軸歸一化，其命名規則爲
bn_name_base + '2c' -
沒有激活函數
-
-
捷徑：
-
此卷積層有F3F3個過濾器，其維度爲（11，11），步伐爲（ss，ss），使用「valid」的填充方式，命名規則爲
conv_name_base + '1' -
此規範層是通道的軸歸一化，其命名規則爲
bn_name_base + '1'
-
-
最後一步：
-
將捷徑與輸入加在一塊兒
-
使用ReLU激活函數
-
咱們要作的是實現卷積塊，請務必查看下面的中文手冊：
def convolutional_block(X, f, filters, stage, block, s=2): """ 實現圖5的卷積塊 參數： X - 輸入的tensor類型的變量，維度爲( m, n_H_prev, n_W_prev, n_C_prev) f - 整數，指定主路徑中間的CONV窗口的維度 filters - 整數列表，定義了主路徑每層的卷積層的過濾器數量 stage - 整數，根據每層的位置來命名每一層，與block參數一塊兒使用。 block - 字符串，據每層的位置來命名每一層，與stage參數一塊兒使用。 s - 整數，指定要使用的步幅 返回： X - 卷積塊的輸出，tensor類型，維度爲(n_H, n_W, n_C) """ #定義命名規則 conv_name_base = "res" + str(stage) + block + "_branch" bn_name_base = "bn" + str(stage) + block + "_branch" #獲取過濾器數量 F1, F2, F3 = filters #保存輸入數據 X_shortcut = X #主路徑 ##主路徑第一部分 X = Conv2D(filters=F1, kernel_size=(1,1), strides=(s,s), padding="valid", name=conv_name_base+"2a", kernel_initializer=glorot_uniform(seed=0))(X) X = BatchNormalization(axis=3,name=bn_name_base+"2a")(X) X = Activation("relu")(X) ##主路徑第二部分 X = Conv2D(filters=F2, kernel_size=(f,f), strides=(1,1), padding="same", name=conv_name_base+"2b", kernel_initializer=glorot_uniform(seed=0))(X) X = BatchNormalization(axis=3,name=bn_name_base+"2b")(X) X = Activation("relu")(X) ##主路徑第三部分 X = Conv2D(filters=F3, kernel_size=(1,1), strides=(1,1), padding="valid", name=conv_name_base+"2c", kernel_initializer=glorot_uniform(seed=0))(X) X = BatchNormalization(axis=3,name=bn_name_base+"2c")(X) #捷徑 X_shortcut = Conv2D(filters=F3, kernel_size=(1,1), strides=(s,s), padding="valid", name=conv_name_base+"1", kernel_initializer=glorot_uniform(seed=0))(X_shortcut) X_shortcut = BatchNormalization(axis=3,name=bn_name_base+"1")(X_shortcut) #最後一步 X = Add()([X,X_shortcut]) X = Activation("relu")(X) return X
- 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
咱們來測試一下：
tf.reset_default_graph()
with tf.Session() as test: np.random.seed(1) A_prev = tf.placeholder("float",[3,4,4,6]) X = np.random.randn(3,4,4,6) A = convolutional_block(A_prev,f=2,filters=[2,4,6],stage=1,block="a") test.run(tf.global_variables_initializer()) out = test.run([A],feed_dict={A_prev:X,K.learning_phase():0}) print("out = " + str(out[0][1][1][0])) test.close()
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10
- 11
- 12
- 13
- 14
測試結果：
out = [ 0.09018463 1.23489773 0.46822017 0.0367176 0. 0.65516603]
- 1
2.3 - 構建你的第一個殘差網絡（50層）
咱們已經作完所須要的全部殘差塊了，下面這個圖就描述了神經網絡的算法細節，圖中的」ID BLOCK「是指標準的恆等塊，」ID BLOCK X3「是指把三個恆等塊放在一塊兒。
這個50層的網絡的細節以下:
1. 對輸入數據進行0填充，padding =（3,3）
-
stage1：
-
卷積層有64個過濾器，其維度爲（7，7），步伐爲（2，2），命名爲「conv1」
-
規範層（BatchNorm）對輸入數據進行通道軸歸一化。
-
最大值池化層使用一個（3，3）的窗口和（2，2）的步伐。
-
-
stage2：
-
卷積塊使用f=3個大小爲[64，64，256]的過濾器，f=3，s=1,block=」a」
-
2個恆等塊使用三個大小爲[64，64，256]的過濾器，f=3，block=」b」、」c」
-
-
stage3：
-
卷積塊使用f=3個大小爲[128,128,512]的過濾器，f=3，s=2,block=」a」
-
3個恆等塊使用三個大小爲[128,128,512]的過濾器，f=3，block=」b」、」c」、」d」
-
-
stage4：
-
卷積塊使用f=3個大小爲[256,256,1024]的過濾器，f=3，s=2,block=」a」
-
5個恆等塊使用三個大小爲[256,256,1024]的過濾器，f=3，block=」b」、」c」、」d」、」e」、」f」
-
-
stage5：
-
卷積塊使用f=3個大小爲[512,512,2048]的過濾器，f=3，s=2,block=」a」
-
2個恆等塊使用三個大小爲[256,256,2048]的過濾器，f=3，block=」b」、」c」
-
-
均值池化層使用維度爲（2,2）的窗口，命名爲「avg_pool」
- 展開操做沒有任何超參數以及命名
- 全鏈接層（密集鏈接）使用softmax激活函數，命名爲
"fc" + str(classes)
爲了實現這50層的殘差網絡，咱們須要查看一下手冊：
def ResNet50(input_shape=(64,64,3),classes=6): """ 實現ResNet50 CONV2D -> BATCHNORM -> RELU -> MAXPOOL -> CONVBLOCK -> IDBLOCK*2 -> CONVBLOCK -> IDBLOCK*3 -> CONVBLOCK -> IDBLOCK*5 -> CONVBLOCK -> IDBLOCK*2 -> AVGPOOL -> TOPLAYER 參數： input_shape - 圖像數據集的維度 classes - 整數，分類數 返回： model - Keras框架的模型 """ #定義tensor類型的輸入數據 X_input = Input(input_shape) #0填充 X = ZeroPadding2D((3,3))(X_input) #stage1 X = Conv2D(filters=64, kernel_size=(7,7), strides=(2,2), name="conv1", kernel_initializer=glorot_uniform(seed=0))(X) X = BatchNormalization(axis=3, name="bn_conv1")(X) X = Activation("relu")(X) X = MaxPooling2D(pool_size=(3,3), strides=(2,2))(X) #stage2 X = convolutional_block(X, f=3, filters=[64,64,256], stage=2, block="a", s=1) X = identity_block(X, f=3, filters=[64,64,256], stage=2, block="b") X = identity_block(X, f=3, filters=[64,64,256], stage=2, block="c") #stage3 X = convolutional_block(X, f=3, filters=[128,128,512], stage=3, block="a", s=2) X = identity_block(X, f=3, filters=[128,128,512], stage=3, block="b") X = identity_block(X, f=3, filters=[128,128,512], stage=3, block="c") X = identity_block(X, f=3, filters=[128,128,512], stage=3, block="d") #stage4 X = convolutional_block(X, f=3, filters=[256,256,1024], stage=4, block="a", s=2) X = identity_block(X, f=3, filters=[256,256,1024], stage=4, block="b") X = identity_block(X, f=3, filters=[256,256,1024], stage=4, block="c") X = identity_block(X, f=3, filters=[256,256,1024], stage=4, block="d") X = identity_block(X, f=3, filters=[256,256,1024], stage=4, block="e") X = identity_block(X, f=3, filters=[256,256,1024], stage=4, block="f") #stage5 X = convolutional_block(X, f=3, filters=[512,512,2048], stage=5, block="a", s=2) X = identity_block(X, f=3, filters=[512,512,2048], stage=5, block="b") X = identity_block(X, f=3, filters=[512,512,2048], stage=5, block="c") #均值池化層 X = AveragePooling2D(pool_size=(2,2),padding="same")(X) #輸出層 X = Flatten()(X) X = Dense(classes, activation="softmax", name="fc"+str(classes), kernel_initializer=glorot_uniform(seed=0))(X) #建立模型 model = Model(inputs=X_input, outputs=X, name="ResNet50") return model
- 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
而後咱們對模型作實體化和編譯工做：
model = ResNet50(input_shape=(64,64,3),classes=6) model.compile(optimizer="adam", loss="categorical_crossentropy", metrics=["accuracy"])
- 1
- 2
- 3
如今模型已經準備好了，接下來就是加載訓練集進行訓練。
X_train_orig, Y_train_orig, X_test_orig, Y_test_orig, classes = resnets_utils.load_dataset()
# Normalize image vectors X_train = X_train_orig / 255. X_test = X_test_orig / 255. # Convert training and test labels to one hot matrices Y_train = resnets_utils.convert_to_one_hot(Y_train_orig, 6).T Y_test = resnets_utils.convert_to_one_hot(Y_test_orig, 6).T print("number of training examples = " + str(X_train.shape[0])) print("number of test examples = " + str(X_test.shape[0])) print("X_train shape: " + str(X_train.shape)) print("Y_train shape: " + str(Y_train.shape)) print("X_test shape: " + str(X_test.shape)) print("Y_test shape: " + str(Y_test.shape))
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10
- 11
- 12
- 13
- 14
- 15
- 16
執行結果：
number of training examples = 1080 number of test examples = 120 X_train shape: (1080, 64, 64, 3) Y_train shape: (1080, 6) X_test shape: (120, 64, 64, 3) Y_test shape: (120, 6)
- 1
- 2
- 3
- 4
- 5
- 6
運行模型兩代，batch=32，每代大約3分鐘左右。
model.fit(X_train,Y_train,epochs=2,batch_size=32)
- 1
執行結果：
Epoch 1/2 1080/1080 [==============================] - 200s 185ms/step - loss: 3.0667 - acc: 0.2593 Epoch 2/2 1080/1080 [==============================] - 186s 172ms/step - loss: 1.9755 - acc: 0.4093
- 1
- 2
- 3
- 4
- 在1212Epoch中，loss在1~5之間算正常，acc在0.2~0.5之間算正常，你的結果和個人不同也算正常。
- 在2222Epoch中，loss在1~5之間算正常，acc在0.2~0.5之間算正常，你能夠看到損失在降低，準確率在上升。
咱們來評估一下模型：
preds = model.evaluate(X_test,Y_test) print("偏差值 = " + str(preds[0])) print("準確率 = " + str(preds[1]))
- 1
- 2
- 3
- 4
- 5
執行結果：
120/120 [==============================] - 5s 44ms/step 偏差值 = 12.3403865178 準確率 = 0.175000000497
- 1
- 2
- 3
在完成這個任務以後，若是願意的話，您還能夠選擇繼續訓練RESNET。當咱們訓練20代時，咱們獲得了更好的性能，可是在得在CPU上訓練須要一個多小時。使用GPU的話，博主已經在手勢數據集上訓練了本身的RESNET50模型的權重，你可使用下面的代碼載並運行博主的訓練模型，加載模型可能須要1min。
#加載模型 model = load_model("ResNet50.h5")
- 1
- 2
而後測試一下博主訓練出來的權值：
preds = model.evaluate(X_test,Y_test) print("偏差值 = " + str(preds[0])) print("準確率 = " + str(preds[1]))
- 1
- 2
- 3
測試結果：
120/120 [==============================] - 4s 35ms/step 偏差值 = 0.108543064694 準確率 = 0.966666662693
- 1
- 2
- 3
2.2.4 使用本身的圖片作測試
按理來講，訓練數據集與本身的數據集是不同的，可是咱們也能夠來試試嘛。
from PIL import Image import numpy as np import matplotlib.pyplot as plt # plt 用於顯示圖片 %matplotlib inline img_path = 'images/fingers_big/2.jpg' my_image = image.load_img(img_path, target_size=(64, 64)) my_image = image.img_to_array(my_image) my_image = np.expand_dims(my_image,axis=0) my_image = preprocess_input(my_image) print("my_image.shape = " + str(my_image.shape)) print("class prediction vector [p(0), p(1), p(2), p(3), p(4), p(5)] = ") print(model.predict(my_image)) my_image = scipy.misc.imread(img_path) plt.imshow(my_image)
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10
- 11
- 12
- 13
- 14
- 15
- 16
- 17
- 18
- 19
- 20
- 21
執行結果：
my_image.shape = (1, 64, 64, 3) class prediction vector [p(0), p(1), p(2), p(3), p(4), p(5)] = [[ 1. 0. 0. 0. 0. 0.]]
- 1
- 2
- 3
請忽略博主這臃腫的手【手動捂臉】
咱們能夠看一下網絡的節點的大小細節：
model.summary()
- 1
執行結果：
__________________________________________________________________________________________________ Layer (type) Output Shape Param # Connected to ================================================================================================== input_1 (InputLayer) (None, 64, 64, 3) 0 __________________________________________________________________________________________________ zero_padding2d_1 (ZeroPadding2D (None, 70, 70, 3) 0 input_1[0][0] __________________________________________________________________________________________________ conv1 (Conv2D) (None, 32, 32, 64) 9472 zero_padding2d_1[0][0] __________________________________________________________________________________________________ bn_conv1 (BatchNormalization) (None, 32, 32, 64) 256 conv1[0][0] __________________________________________________________________________________________________ activation_1 (Activation) (None, 32, 32, 64) 0 bn_conv1[0][0] __________________________________________________________________________________________________ max_pooling2d_1 (MaxPooling2D) (None, 15, 15, 64) 0 activation_1[0][0] __________________________________________________________________________________________________ res2a_branch2a (Conv2D) (None, 15, 15, 64) 4160 max_pooling2d_1[0][0] __________________________________________________________________________________________________ bn2a_branch2a (BatchNormalizati (None, 15, 15, 64) 256 res2a_branch2a[0][0] __________________________________________________________________________________________________ activation_2 (Activation) (None, 15, 15, 64) 0 bn2a_branch2a[0][0] __________________________________________________________________________________________________ res2a_branch2b (Conv2D) (None, 15, 15, 64) 36928 activation_2[0][0] __________________________________________________________________________________________