Image Classification¶
In this project, you'll classify images from the CIFAR-10 dataset. The dataset consists of airplanes, dogs, cats, and other objects. You'll preprocess the images, then train a convolutional neural network on all the samples. The images need to be normalized and the labels need to be one-hot encoded. You'll get to apply what you learned and build a convolutional, max pooling, dropout, and fully connected layers. At the end, you'll get to see your neural network's predictions on the sample images.
Get the Data¶
Run the following cell to download the CIFAR-10 dataset for python.
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
from urllib.request import urlretrieve
from os.path import isfile, isdir
from tqdm import tqdm
import problem_unittests as tests
import tarfile
cifar10_dataset_folder_path = 'cifar-10-batches-py'
class DLProgress(tqdm):
last_block = 0
def hook(self, block_num=1, block_size=1, total_size=None):
self.total = total_size
self.update((block_num - self.last_block) * block_size)
self.last_block = block_num
if not isfile('cifar-10-python.tar.gz'):
with DLProgress(unit='B', unit_scale=True, miniters=1, desc='CIFAR-10 Dataset') as pbar:
urlretrieve(
'https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz',
'cifar-10-python.tar.gz',
pbar.hook)
if not isdir(cifar10_dataset_folder_path):
with tarfile.open('cifar-10-python.tar.gz') as tar:
tar.extractall()
tar.close()
tests.test_folder_path(cifar10_dataset_folder_path)
Explore the Data¶
The dataset is broken into batches to prevent your machine from running out of memory. The CIFAR-10 dataset consists of 5 batches, named data_batch_1, data_batch_2, etc.. Each batch contains the labels and images that are one of the following:
- airplane
- automobile
- bird
- cat
- deer
- dog
- frog
- horse
- ship
- truck
Understanding a dataset is part of making predictions on the data. Play around with the code cell below by changing the batch_id and sample_id. The batch_id is the id for a batch (1-5). The sample_id is the id for a image and label pair in the batch.
Ask yourself "What are all possible labels?", "What is the range of values for the image data?", "Are the labels in order or random?". Answers to questions like these will help you preprocess the data and end up with better predictions.
%matplotlib inline
%config InlineBackend.figure_format = 'retina'
import helper
import numpy as np
# Explore the dataset
batch_id = 2
sample_id = 9999
helper.display_stats(cifar10_dataset_folder_path, batch_id, sample_id)
Answer to the data exploration questions¶
There are 10 possible labels: airplane, automobile, bird etc. These labels are encoded to 0, 1, 2, ..., 9. Please note that the encoding starts from 0, not 1. The batch id starts from 1, and ends at 5; the image id starts from 0, and ends at 9999. So there are totally 10K images per batch, and 50K images for all the batches.
Each image contains values ranges from 0 to 255, the dimension is 32(width)x32(height)x3(color). The labels are in random order.
def normalize(x):
"""
Normalize a list of sample image data in the range of 0 to 1
: x: List of image data. The image shape is (32, 32, 3)
: return: Numpy array of normalize data
"""
# TODO: Implement Function
# return None
a = 0
b = 255
return (x-a)/(b-a)
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_normalize(normalize)
One-hot encode¶
Just like the previous code cell, you'll be implementing a function for preprocessing. This time, you'll implement the one_hot_encode function. The input, x, are a list of labels. Implement the function to return the list of labels as One-Hot encoded Numpy array. The possible values for labels are 0 to 9. The one-hot encoding function should return the same encoding for each value between each call to one_hot_encode. Make sure to save the map of encodings outside the function.
Hint: Don't reinvent the wheel.
def one_hot_encode(x):
"""
One hot encode a list of sample labels. Return a one-hot encoded vector for each label.
: x: List of sample Labels
: return: Numpy array of one-hot encoded labels
"""
# TODO: Implement Function
# return None
y = np.zeros((len(x), 10))
for i in range(len(x)):
y[i,x[i]] = 1
return y
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_one_hot_encode(one_hot_encode)
Randomize Data¶
As you saw from exploring the data above, the order of the samples are randomized. It doesn't hurt to randomize it again, but you don't need to for this dataset.
Preprocess all the data and save it¶
Running the code cell below will preprocess all the CIFAR-10 data and save it to file. The code below also uses 10% of the training data for validation.
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
# Preprocess Training, Validation, and Testing Data
helper.preprocess_and_save_data(cifar10_dataset_folder_path, normalize, one_hot_encode)
Check Point¶
This is your first checkpoint. If you ever decide to come back to this notebook or have to restart the notebook, you can start from here. The preprocessed data has been saved to disk.
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
import pickle
import problem_unittests as tests
import helper
# Load the Preprocessed Validation data
valid_features, valid_labels = pickle.load(open('preprocess_validation.p', mode='rb'))
#need to delete
print(valid_labels)
valid_labels.shape
valid_features.shape
valid_features
Build the network¶
For the neural network, you'll build each layer into a function. Most of the code you've seen has been outside of functions. To test your code more thoroughly, we require that you put each layer in a function. This allows us to give you better feedback and test for simple mistakes using our unittests before you submit your project.
Note: If you're finding it hard to dedicate enough time for this course each week, we've provided a small shortcut to this part of the project. In the next couple of problems, you'll have the option to use classes from the TensorFlow Layers or TensorFlow Layers (contrib) packages to build each layer, except the layers you build in the "Convolutional and Max Pooling Layer" section. TF Layers is similar to Keras's and TFLearn's abstraction to layers, so it's easy to pickup.
However, if you would like to get the most out of this course, try to solve all the problems without using anything from the TF Layers packages. You can still use classes from other packages that happen to have the same name as ones you find in TF Layers! For example, instead of using the TF Layers version of the
conv2dclass, tf.layers.conv2d, you would want to use the TF Neural Network version ofconv2d, tf.nn.conv2d.
Let's begin!
Input¶
The neural network needs to read the image data, one-hot encoded labels, and dropout keep probability. Implement the following functions
- Implement
neural_net_image_input- Return a TF Placeholder
- Set the shape using
image_shapewith batch size set toNone. - Name the TensorFlow placeholder "x" using the TensorFlow
nameparameter in the TF Placeholder.
- Implement
neural_net_label_input- Return a TF Placeholder
- Set the shape using
n_classeswith batch size set toNone. - Name the TensorFlow placeholder "y" using the TensorFlow
nameparameter in the TF Placeholder.
- Implement
neural_net_keep_prob_input- Return a TF Placeholder for dropout keep probability.
- Name the TensorFlow placeholder "keep_prob" using the TensorFlow
nameparameter in the TF Placeholder.
These names will be used at the end of the project to load your saved model.
Note: None for shapes in TensorFlow allow for a dynamic size.
import tensorflow as tf
def neural_net_image_input(image_shape):
"""
Return a Tensor for a bach of image input
: image_shape: Shape of the images
: return: Tensor for image input.
"""
# TODO: Implement Function
return tf.placeholder(tf.float32, [None] + list(image_shape), "x")
def neural_net_label_input(n_classes):
"""
Return a Tensor for a batch of label input
: n_classes: Number of classes
: return: Tensor for label input.
"""
# TODO: Implement Function
return tf.placeholder(tf.float32, [None, n_classes], "y")
def neural_net_keep_prob_input():
"""
Return a Tensor for keep probability
: return: Tensor for keep probability.
"""
# TODO: Implement Function
return tf.placeholder(tf.float32, None, "keep_prob")
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tf.reset_default_graph()
tests.test_nn_image_inputs(neural_net_image_input)
tests.test_nn_label_inputs(neural_net_label_input)
tests.test_nn_keep_prob_inputs(neural_net_keep_prob_input)
Convolution and Max Pooling Layer¶
Convolution layers have a lot of success with images. For this code cell, you should implement the function conv2d_maxpool to apply convolution then max pooling:
- Create the weight and bias using
conv_ksize,conv_num_outputsand the shape ofx_tensor. - Apply a convolution to
x_tensorusing weight andconv_strides.- We recommend you use same padding, but you're welcome to use any padding.
- Add bias
- Add a nonlinear activation to the convolution.
- Apply Max Pooling using
pool_ksizeandpool_strides.- We recommend you use same padding, but you're welcome to use any padding.
Note: You can't use TensorFlow Layers or TensorFlow Layers (contrib) for this layer, but you can still use TensorFlow's Neural Network package. You may still use the shortcut option for all the other layers.
#need to delete
conv_ksize=(32,32)
y=conv_ksize+(3,)+(5,)
print(y)
x=tf.truncated_normal(y)
print(x)
def conv2d_maxpool(x_tensor, conv_num_outputs, conv_ksize, conv_strides, pool_ksize, pool_strides):
"""
Apply convolution then max pooling to x_tensor
:param x_tensor: TensorFlow Tensor
:param conv_num_outputs: Number of outputs for the convolutional layer
:param conv_ksize: kernal size 2-D Tuple for the convolutional layer
:param conv_strides: Stride 2-D Tuple for convolution
:param pool_ksize: kernal size 2-D Tuple for pool
:param pool_strides: Stride 2-D Tuple for pool
: return: A tensor that represents convolution and max pooling of x_tensor
"""
# TODO: Implement Function
dimension = x_tensor.get_shape().as_list()
shape = list(conv_ksize + (dimension[-1],) + (conv_num_outputs,))
#print(shape)
filter_weights = tf.Variable(tf.truncated_normal(shape,0,0.1)) # (height, width, input_depth, output_depth)
filter_bias = tf.Variable(tf.zeros(conv_num_outputs))
padding = 'SAME'
#print(list((1,)+conv_strides+(1,)))
#print(filter_weights)
conv_layer = tf.nn.conv2d(x_tensor, filter_weights, list((1,)+conv_strides+(1,)), padding)
conv_layer = tf.nn.bias_add(conv_layer, filter_bias)
conv_layer = tf.nn.relu(conv_layer)
conv_layer = tf.nn.max_pool(
conv_layer,
ksize=[1] + list(pool_ksize) + [1],
strides=[1] + list(pool_strides) + [1],
padding='SAME')
return conv_layer
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_con_pool(conv2d_maxpool)
Flatten Layer¶
Implement the flatten function to change the dimension of x_tensor from a 4-D tensor to a 2-D tensor. The output should be the shape (Batch Size, Flattened Image Size). Shortcut option: you can use classes from the TensorFlow Layers or TensorFlow Layers (contrib) packages for this layer. For more of a challenge, only use other TensorFlow packages.
from numpy import prod
def flatten(x_tensor):
"""
Flatten x_tensor to (Batch Size, Flattened Image Size)
: x_tensor: A tensor of size (Batch Size, ...), where ... are the image dimensions.
: return: A tensor of size (Batch Size, Flattened Image Size).
"""
# TODO: Implement Function
#return None
dimension = x_tensor.get_shape().as_list()
return tf.reshape(x_tensor,[-1,prod(dimension[1:])])
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_flatten(flatten)
Fully-Connected Layer¶
Implement the fully_conn function to apply a fully connected layer to x_tensor with the shape (Batch Size, num_outputs). Shortcut option: you can use classes from the TensorFlow Layers or TensorFlow Layers (contrib) packages for this layer. For more of a challenge, only use other TensorFlow packages.
def fully_conn(x_tensor, num_outputs):
"""
Apply a fully connected layer to x_tensor using weight and bias
: x_tensor: A 2-D tensor where the first dimension is batch size.
: num_outputs: The number of output that the new tensor should be.
: return: A 2-D tensor where the second dimension is num_outputs.
"""
# TODO: Implement Function
#return None
dimension = x_tensor.get_shape().as_list()
shape = list( (dimension[-1],) + (num_outputs,))
#print(shape)
weight = tf.Variable(tf.truncated_normal(shape,0,0.1))
bias = tf.Variable(tf.zeros(num_outputs))
return tf.nn.relu(tf.add(tf.matmul(x_tensor,weight), bias))
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_fully_conn(fully_conn)
Output Layer¶
Implement the output function to apply a fully connected layer to x_tensor with the shape (Batch Size, num_outputs). Shortcut option: you can use classes from the TensorFlow Layers or TensorFlow Layers (contrib) packages for this layer. For more of a challenge, only use other TensorFlow packages.
Note: Activation, softmax, or cross entropy should not be applied to this.
def output(x_tensor, num_outputs):
"""
Apply a output layer to x_tensor using weight and bias
: x_tensor: A 2-D tensor where the first dimension is batch size.
: num_outputs: The number of output that the new tensor should be.
: return: A 2-D tensor where the second dimension is num_outputs.
"""
# TODO: Implement Function
# return None
dimension = x_tensor.get_shape().as_list()
shape = list( (dimension[-1],) + (num_outputs,))
#print(shape)
weight = tf.Variable(tf.truncated_normal(shape,0,0.01))
bias = tf.Variable(tf.zeros(num_outputs))
return tf.add(tf.matmul(x_tensor,weight), bias)
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_output(output)
Create Convolutional Model¶
Implement the function conv_net to create a convolutional neural network model. The function takes in a batch of images, x, and outputs logits. Use the layers you created above to create this model:
- Apply 1, 2, or 3 Convolution and Max Pool layers
- Apply a Flatten Layer
- Apply 1, 2, or 3 Fully Connected Layers
- Apply an Output Layer
- Return the output
- Apply TensorFlow's Dropout to one or more layers in the model using
keep_prob.
def conv_net(x, keep_prob):
"""
Create a convolutional neural network model
: x: Placeholder tensor that holds image data.
: keep_prob: Placeholder tensor that hold dropout keep probability.
: return: Tensor that represents logits
"""
# TODO: Apply 1, 2, or 3 Convolution and Max Pool layers
# Play around with different number of outputs, kernel size and stride
# Function Definition from Above:
# conv2d_maxpool(x_tensor, conv_num_outputs, conv_ksize, conv_strides, pool_ksize, pool_strides)
model = conv2d_maxpool(x, conv_num_outputs=18, conv_ksize=(4,4), conv_strides=(1,1), pool_ksize=(8,8), pool_strides=(1,1))
#model = conv2d_maxpool(x, conv_num_outputs=36, conv_ksize=(4,4), conv_strides=(1,1), pool_ksize=(2,2), pool_strides=(1,1))
#model = conv2d_maxpool(model, conv_num_outputs=8, conv_ksize=(4,4), conv_strides=(1,1), pool_ksize=(4,4), pool_strides=(1,1))
#model = tf.nn.relu(model)
model = tf.nn.dropout(model, keep_prob)
# TODO: Apply a Flatten Layer
# Function Definition from Above:
# flatten(x_tensor)
model = flatten(model)
# TODO: Apply 1, 2, or 3 Fully Connected Layers
# Play around with different number of outputs
# Function Definition from Above:
# fully_conn(x_tensor, num_outputs)
model = fully_conn(model,384)
#model = fully_conn(model,200)
#model = fully_conn(model,20)
#model = tf.nn.relu(model)
model = tf.nn.dropout(model, keep_prob)
# TODO: Apply an Output Layer
# Set this to the number of classes
# Function Definition from Above:
# output(x_tensor, num_outputs)
model = output(model,10)
# TODO: return output
return model
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
##############################
## Build the Neural Network ##
##############################
# Remove previous weights, bias, inputs, etc..
tf.reset_default_graph()
# Inputs
x = neural_net_image_input((32, 32, 3))
y = neural_net_label_input(10)
keep_prob = neural_net_keep_prob_input()
# Model
logits = conv_net(x, keep_prob)
# Name logits Tensor, so that is can be loaded from disk after training
logits = tf.identity(logits, name='logits')
# Loss and Optimizer
cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=logits, labels=y))
optimizer = tf.train.AdamOptimizer().minimize(cost)
# Accuracy
correct_pred = tf.equal(tf.argmax(logits, 1), tf.argmax(y, 1))
accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32), name='accuracy')
tests.test_conv_net(conv_net)
Train the Neural Network¶
Single Optimization¶
Implement the function train_neural_network to do a single optimization. The optimization should use optimizer to optimize in session with a feed_dict of the following:
xfor image inputyfor labelskeep_probfor keep probability for dropout
This function will be called for each batch, so tf.global_variables_initializer() has already been called.
Note: Nothing needs to be returned. This function is only optimizing the neural network.
def train_neural_network(session, optimizer, keep_probability, feature_batch, label_batch):
"""
Optimize the session on a batch of images and labels
: session: Current TensorFlow session
: optimizer: TensorFlow optimizer function
: keep_probability: keep probability
: feature_batch: Batch of Numpy image data
: label_batch: Batch of Numpy label data
"""
# TODO: Implement Function
# pass
session.run(optimizer, feed_dict={x:feature_batch, y:label_batch, keep_prob:keep_probability})
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_train_nn(train_neural_network)
Show Stats¶
Implement the function print_stats to print loss and validation accuracy. Use the global variables valid_features and valid_labels to calculate validation accuracy. Use a keep probability of 1.0 to calculate the loss and validation accuracy.
def print_stats(session, feature_batch, label_batch, cost, accuracy):
"""
Print information about loss and validation accuracy
: session: Current TensorFlow session
: feature_batch: Batch of Numpy image data
: label_batch: Batch of Numpy label data
: cost: TensorFlow cost function
: accuracy: TensorFlow accuracy function
"""
# TODO: Implement Function
# pass
loss = session.run(cost, feed_dict={x:feature_batch, y:label_batch, keep_prob:1.0})
valid_acc = sess.run(accuracy, feed_dict={
x: valid_features,
y: valid_labels,
keep_prob: 1.})
print('Loss: {:>10.4f} Validation Accuracy: {:.6f}'.format(
loss,
valid_acc))
Hyperparameters¶
Tune the following parameters:
- Set
epochsto the number of iterations until the network stops learning or start overfitting - Set
batch_sizeto the highest number that your machine has memory for. Most people set them to common sizes of memory:- 64
- 128
- 256
- ...
- Set
keep_probabilityto the probability of keeping a node using dropout
# TODO: Tune Parameters
epochs = 100
batch_size = 512
keep_probability = 0.3
Train on a Single CIFAR-10 Batch¶
Instead of training the neural network on all the CIFAR-10 batches of data, let's use a single batch. This should save time while you iterate on the model to get a better accuracy. Once the final validation accuracy is 50% or greater, run the model on all the data in the next section.
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
print('Checking the Training on a Single Batch...')
with tf.Session() as sess:
# Initializing the variables
sess.run(tf.global_variables_initializer())
# Training cycle
for epoch in range(epochs):
batch_i = 1
for batch_features, batch_labels in helper.load_preprocess_training_batch(batch_i, batch_size):
train_neural_network(sess, optimizer, keep_probability, batch_features, batch_labels)
print('Epoch {:>2}, CIFAR-10 Batch {}: '.format(epoch + 1, batch_i), end='')
print_stats(sess, batch_features, batch_labels, cost, accuracy)
Fully Train the Model¶
Now that you got a good accuracy with a single CIFAR-10 batch, try it with all five batches.
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
save_model_path = './image_classification'
print('Training...')
with tf.Session() as sess:
# Initializing the variables
sess.run(tf.global_variables_initializer())
# Training cycle
for epoch in range(epochs):
# Loop over all batches
n_batches = 5
for batch_i in range(1, n_batches + 1):
for batch_features, batch_labels in helper.load_preprocess_training_batch(batch_i, batch_size):
train_neural_network(sess, optimizer, keep_probability, batch_features, batch_labels)
print('Epoch {:>2}, CIFAR-10 Batch {}: '.format(epoch + 1, batch_i), end='')
print_stats(sess, batch_features, batch_labels, cost, accuracy)
# Save Model
saver = tf.train.Saver()
save_path = saver.save(sess, save_model_path)
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
%matplotlib inline
%config InlineBackend.figure_format = 'retina'
import tensorflow as tf
import pickle
import helper
import random
# Set batch size if not already set
try:
if batch_size:
pass
except NameError:
batch_size = 64
save_model_path = './image_classification'
n_samples = 4
top_n_predictions = 3
def test_model():
"""
Test the saved model against the test dataset
"""
test_features, test_labels = pickle.load(open('preprocess_training.p', mode='rb'))
loaded_graph = tf.Graph()
with tf.Session(graph=loaded_graph) as sess:
# Load model
loader = tf.train.import_meta_graph(save_model_path + '.meta')
loader.restore(sess, save_model_path)
# Get Tensors from loaded model
loaded_x = loaded_graph.get_tensor_by_name('x:0')
loaded_y = loaded_graph.get_tensor_by_name('y:0')
loaded_keep_prob = loaded_graph.get_tensor_by_name('keep_prob:0')
loaded_logits = loaded_graph.get_tensor_by_name('logits:0')
loaded_acc = loaded_graph.get_tensor_by_name('accuracy:0')
# Get accuracy in batches for memory limitations
test_batch_acc_total = 0
test_batch_count = 0
for train_feature_batch, train_label_batch in helper.batch_features_labels(test_features, test_labels, batch_size):
test_batch_acc_total += sess.run(
loaded_acc,
feed_dict={loaded_x: train_feature_batch, loaded_y: train_label_batch, loaded_keep_prob: 1.0})
test_batch_count += 1
print('Testing Accuracy: {}\n'.format(test_batch_acc_total/test_batch_count))
# Print Random Samples
random_test_features, random_test_labels = tuple(zip(*random.sample(list(zip(test_features, test_labels)), n_samples)))
random_test_predictions = sess.run(
tf.nn.top_k(tf.nn.softmax(loaded_logits), top_n_predictions),
feed_dict={loaded_x: random_test_features, loaded_y: random_test_labels, loaded_keep_prob: 1.0})
helper.display_image_predictions(random_test_features, random_test_labels, random_test_predictions)
test_model()
Why 50-70% Accuracy?¶
You might be wondering why you can't get an accuracy any higher. First things first, 50% isn't bad for a simple CNN. Pure guessing would get you 10% accuracy. However, you might notice people are getting scores well above 70%. That's because we haven't taught you all there is to know about neural networks. We still need to cover a few more techniques.
Submitting This Project¶
When submitting this project, make sure to run all the cells before saving the notebook. Save the notebook file as "dlnd_image_classification.ipynb" and save it as a HTML file under "File" -> "Download as". Include the "helper.py" and "problem_unittests.py" files in your submission.
Some thoughts¶
In this project, I find parameter tuning is an important and time-consuming step. Building the network per se is a relatively easy job, especially if we use tensorflow or other libraries. However, chooseing the right parameter to ensure a high accuracy is hard.
In the beginning, I found no matter what I do, I could not achieve a 40% accuracy. After reading forum discussions, I noticed that I have initialized the weights using the default paremter (standard deviation 1). When I chose a much smaller value (standard deviation = 0.1), the accuracy jumped to above 0.55.
The keep_probability also matters. I find lower keep_probability gives a better result (though the process is slower). After trying a few values, including 0.5, 0.1, 0.3, etc, I eventually chose 0.3.
The number of output in the convolution layer is set to 18, which is larger than my initial value. The ksize of the max_pooling layer was set to 4 initially, but after changing to 8 the accuracy increased to 0.64 from 0.61.
The final accuracy of the single-batch training is 0.64, for all the 5 batches it is 0.71, and for the test data it is 0.71. This is an acceptable result.
The reviewer suggested to apply ReLU for the convolution layer (before max pooling), and for the fully connected layer. In the first submission ReLU was applied in the convolutional model. I moved ReLU inside the convolutional layer and fully connected layer definition. The final accuracy is largely unchange (0.70).