[Tensorflow2.X][轉載]深度可卷積神經網絡實現mnist識別

from tensorflow import keras

import tensorflow as tf

import matplotlib.pyplot as plt

import numpy as np

import pandas as pd

import sklearn

import os

import sys

 

def plot_learning_curves(history):

 

    pd.DataFrame(history.history).plot(figsize=(8,5))

    plt.grid(True)

    plt.gca().set_ylim(0,1)

    plt.show()

 

 

fashion_mnist = keras.datasets.fashion_mnist

(x_train_all, y_train_all),(x_test,y_test) = fashion_mnist.load_data()

x_valid,x_train = x_train_all[:5000],x_train_all[5000:]

y_valid,y_train = y_train_all[:5000],y_train_all[5000:]

 

#數據歸一化

from sklearn.preprocessing import StandardScaler

 

scaler = StandardScaler()

#x_train: [None, 28, 28] -> [None,784]

x_train_scaled = scaler.fit_transform(x_train.astype(np.float32).reshape(-1, 1)).reshape(-1, 28, 28,1)

x_valid_scaled = scaler.transform(x_valid.astype(np.float32).reshape(-1,1)).reshape(-1,28,28,1)

x_test_scaled = scaler.transform(x_test.astype(np.float32).reshape(-1,1)).reshape(-1,28,28,1)

 

 

 

#使用relu函數

# model = tf.keras.models.Sequential()

# model.add(keras.layers.Conv2D(filter=32,kernel_size=3,

#                               padding='same',activation='relu',

#                               input_shape=(28,28,1)))

#

# model.add(keras.layers.Conv2D(filters=32,kernel_size=3,

#                               padding='same',activation='relu'))

# model.add(keras.layers.MaxPooling2D(pool_size=2))

#

#

# model.add(keras.layers.Conv2D(filter=64, kernel_size=3,

#                               padding='same',activation='relu'))

#

# model.add(keras.layers.Conv2D(filters=64,kernel_size=3,

#                               padding='same',activation='relu'))

# model.add(keras.layers.MaxPooling2D(pool_size=2))

#

# model.add(keras.layers.Conv2D(filter=128,kernel_size=3,

#                               padding='same',activation='relu'))

#

# model.add(keras.layers.Conv2D(filters=128,kernel_size=3,

#                               padding='same',activation='relu'))

# model.add(keras.layers.MaxPooling2D(pool_size=2))

# model.add(keras.layers.Flatten())

#

# model.add(keras.layers.Dense(128,activation='relu'))

# model.add(keras.layers.Dense(10,activation='softmax'))

 

#使用selu函數

model = tf.keras.models.Sequential()

model.add(keras.layers.Conv2D(filters=32,kernel_size=3,

                              padding='same',activation='selu',

                              input_shape=(28,28,1)))

 

model.add(keras.layers.SeparableConv2D(filters=32,kernel_size=3,

                                        padding='same',activation='selu'))

model.add(keras.layers.MaxPooling2D(pool_size=2))

 

 

model.add(keras.layers.SeparableConv2D(filters=64, kernel_size=3,

                              padding='same',activation='selu'))

 

model.add(keras.layers.SeparableConv2D(filters=64,kernel_size=3,

                              padding='same',activation='selu'))

model.add(keras.layers.MaxPooling2D(pool_size=2))

 

model.add(keras.layers.SeparableConv2D(filters=128,kernel_size=3,

                              padding='same',activation='selu'))

 

model.add(keras.layers.SeparableConv2D(filters=128,kernel_size=3,

                              padding='same',activation='selu'))

model.add(keras.layers.MaxPooling2D(pool_size=2))

model.add(keras.layers.Flatten())

 

model.add(keras.layers.Dense(128,activation='selu'))

#model.add(keras.layers.Dense(10,activation='softmax'))

model.compile(loss='sparse_categorical_crossentropy', optimizer='', metrics=['accuracy'])

 

#使用了三個callback：Tensorboard, earlystopping, ModelCheckpoint

#logdir = './spearable-cnn-selu-callbacks'

logdir = os.path.join("dnn-callbacks")

if not os.path.exists(logdir):

    os.mkdir(logdir)

output_model_file = os.path.join(logdir,"fashion_mnist_model.h5")

print("out:",output_model_file)

callbacks = [

    tf.keras.callbacks.TensorBoard(logdir),

    tf.keras.callbacks.ModelCheckpoint(output_model_file,save_best_only=True),

    tf.keras.callbacks.EarlyStopping(patience=5, min_delta=1e-3)

]

history = model.fit(x_train_scaled, y_train, epochs=10, validation_data=(x_valid_scaled, y_valid),callbacks=callbacks)

 

plot_learning_curves(history)