-
Notifications
You must be signed in to change notification settings - Fork 0
/
vgg_TransferLearning.py
175 lines (139 loc) · 7.72 KB
/
vgg_TransferLearning.py
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
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
import os
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import tensorflow as tf
from tensorflow import keras
from tensorflow.keras import Sequential
from tensorflow.keras.layers import Input, Flatten, Dense, Conv2D, MaxPool2D, ZeroPadding2D, Dropout, BatchNormalization
from tensorflow.keras.preprocessing.image import ImageDataGenerator
from tensorflow.keras import layers
from tensorflow.keras.applications.vgg16 import VGG16
from tensorflow.keras.applications.vgg19 import VGG19
from tensorflow.keras.models import Model
from tensorflow.keras.optimizers import Adam, SGD
from tensorflow.keras.callbacks import EarlyStopping
################################################################################################
#
# This program continues from the program, "vgg16_intro.py" and "vgg16_SaveModel.py".
# This program demonstrates data augmentations, transfer learning and exponentialDecay learning rate.
# We use both vgg16 and vgg19
#
# We also apply early stopping in training
# We can then use "vgg16_LoadModel.py" to load the saved model and predict a single image.
#
################################################################################################
batch_size = 256 # Using Quadro RTX 5000 with 16G video RAM.
# Reduce the batch_size if the vidoe RAM is not enough
img_width = 224 # (224, 224, 3) standard input size for VGG model
img_height = 224
# Download the "flower_photos.tgz" from below website, and put into "date_dir"
# /datasets/flowers_photos/daisy/
# /datasets/flowers_photos/dandelion/
# /datasets/flowers_photos/roses/
# /datasets/flowers_photos/sunflowers/
# /datasets/flowers_photos/tulips/
# https://www.tensorflow.org/tutorials/images/classification
# https://storage.googleapis.com/download.tensorflow.org/example_images/flower_photos.tgz
data_dir = "E:/datasets/flower_photos/"
############################################################################
# Data augmentation
datagen = ImageDataGenerator(rescale = 1./255,
rotation_range=40,
width_shift_range = 0.2,
height_shift_range = 0.2,
shear_range=0.2,
zoom_range=0.2,
horizontal_flip=True,
vertical_flip=True,
validation_split = 0.2) #set validation split
classes = ['daisy', 'dandelion', 'roses', 'sunflowers', 'tulips']
num_classes = len(classes)
train_ds = datagen.flow_from_directory(directory=data_dir,
target_size= (img_width, img_height),
classes = classes,
class_mode = 'categorical',
shuffle = True,
batch_size = batch_size,
subset='training')
val_ds = datagen.flow_from_directory(directory = data_dir,
target_size= (img_width, img_height),
classes = classes,
class_mode = 'categorical',
shuffle = False,
batch_size = batch_size,
subset='validation')
############################################################################
# Use VGG16 with pre-trained weights, and apply transfer learning
def VGG16_v1(input_shape = (224,224,3), classes = num_classes):
model = VGG16(include_top=False, input_shape= input_shape, weights='imagenet')
for layer in model.layers:
layer.trainable = False #15 = True
X = model.layers[-5].output
X = Conv2D(filters=512,kernel_size=(3,3),padding="same", activation="relu", kernel_initializer = 'he_uniform', name = "conv-512a")(X)
X = Conv2D(filters=512,kernel_size=(3,3),padding="same", activation="relu", kernel_initializer = 'he_uniform', name = "conv-512b")(X)
X = Conv2D(filters=512,kernel_size=(3,3),padding="same", activation="relu", kernel_initializer = 'he_uniform', name = "conv-512c")(X)
X = MaxPool2D(pool_size=(2,2),strides=(2,2), name="maxpool-5")(X)
X = Dropout(0.2)(X)
X = Flatten()(X)
X = Dense(512, activation='relu', kernel_initializer='he_uniform', name = 'Dense-512-1')(X)
X= Dropout(0.5)(X)
output = Dense(num_classes, activation='softmax', name = 'output_layer')(X)
model = Model(inputs=model.inputs, outputs=output)
return model
############################################################################
# Use VGG19 with pre-trained weights, and apply transfer learning
def VGG19_v1(input_shape = (224,224,3), classes = num_classes):
model = VGG16(include_top=False, input_shape= input_shape, weights='imagenet')
for layer in model.layers:
layer.trainable = False
X = model.layers[-5].output
X = Conv2D(filters=512,kernel_size=(3,3),padding="same", activation="relu", kernel_initializer = 'he_uniform', name = "conv-512a")(X)
X = Conv2D(filters=512,kernel_size=(3,3),padding="same", activation="relu", kernel_initializer = 'he_uniform', name = "conv-512b")(X)
X = Conv2D(filters=512,kernel_size=(3,3),padding="same", activation="relu", kernel_initializer = 'he_uniform', name = "conv-512c")(X)
X = Conv2D(filters=512,kernel_size=(3,3),padding="same", activation="relu", kernel_initializer = 'he_uniform', name = "conv-512d")(X)
X = MaxPool2D(pool_size=(2,2),strides=(2,2), name="maxpool-5")(X)
X = Dropout(0.2)(X)
X = Flatten()(X)
X = Dense(4096, activation='relu', kernel_initializer='he_uniform', name = 'Dense-4096-1')(X)
X = Dropout(0.5)(X)
X = Dense(4096, activation='relu', kernel_initializer='he_uniform', name = 'Dense-4096-2')(X)
X = Dropout(0.5)(X)
output = Dense(num_classes, activation='softmax', name = 'output_layer')(X)
model = Model(inputs=model.inputs, outputs=output)
return model
#model = VGG16_v1(input_shape = [img_width,img_height,3], classes = num_classes)
model = VGG19_v1(input_shape = [img_width,img_height,3], classes = num_classes)
model.summary()
############################################################################
# Use learnng rate schedule -> exponentialDecay
initial_learning_rate = 0.001
lr_schedule = tf.keras.optimizers.schedules.ExponentialDecay(
initial_learning_rate,
decay_steps=100000,
decay_rate=0.96,
staircase=True)
#opt = SGD(lr=0.00002, momentum=0.9)
opt = SGD(learning_rate = lr_schedule)
model.compile(optimizer=opt, loss='categorical_crossentropy', metrics=['accuracy'])
############################################################################
# Apply early stopping commands
monitor = EarlyStopping(monitor='val_loss', min_delta =0.001, patience = 15, verbose = 0,
mode = 'auto', restore_best_weights = True )
epochs = 250
history = model.fit(train_ds,
steps_per_epoch = len(train_ds), #batch_size
epochs = epochs,
validation_data = val_ds, #batch_size
validation_steps = len(val_ds),
callbacks=[monitor]
)
############################################################################
# Save a model
model.save("vgg19_t1.h5")
############################################################################
# Save history of training
df = pd.DataFrame(history.history)
filename = 'history_vgg19_t1.csv'
with open (filename, mode ='w') as f:
df.to_csv(f)