-
Notifications
You must be signed in to change notification settings - Fork 1
/
train.py
124 lines (108 loc) · 3.78 KB
/
train.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
import tensorflow as tf
from tensorflow import keras
from tensorflow.keras.callbacks import TensorBoard
import numpy as np
import matplotlib.pyplot as plt
import copy
import generate
import time
t = str(int(time.time()))
RUN_NAME = 'Training Set' + t
MODEL_NAME = 'MyTrainedModel'+t+'.ckpt'
G = generate.Gen()
ValidationSet,ValidationLabel = G.generateNumpySet(1000,G.letters, G.fonts)
ValidationSet = ValidationSet/255.0
# Plot labeled Images
plt.figure(figsize=(10,10))
for i in range(25):
plt.subplot(5,5,i+1)
plt.xticks([])
plt.yticks([])
plt.grid(False)
plt.imshow(ValidationSet[i], cmap=plt.cm.binary)
plt.xlabel(G.letters[ValidationLabel[i]])
plt.show() #(But not right now)
model = keras.Sequential()
model.add(keras.layers.Conv2D(64, kernel_size=3, activation='relu', input_shape=(G.size[0], G.size[1], 3)))
model.add(keras.layers.Conv2D(64, kernel_size=3, activation='relu'))
model.add(keras.layers.Flatten())
model.add(keras.layers.Dense(128, activation=tf.nn.relu))
model.add(keras.layers.Dropout(0.1))
model.add(keras.layers.Dense(128, activation=tf.nn.relu))
model.add(keras.layers.Dropout(0.1))
model.add(keras.layers.Dense(len(G.letters), activation=tf.nn.softmax))
model.compile(optimizer='adam',
loss='sparse_categorical_crossentropy',
metrics=['accuracy'])
#Keras Logger
logger = keras.callbacks.TensorBoard(
log_dir = 'logs/' + RUN_NAME,
write_graph = True
)
earlyStop = tf.keras.callbacks.EarlyStopping(monitor='val_loss', patience=2)
model.summary()
set,label = G.generateNumpySet(5000,G.letters, G.fonts)
#Pre-training
test_acc = 0.0
while test_acc < 0.1:
model.fit(set, label,
epochs=10,
callbacks=[logger],
validation_data=(ValidationSet,ValidationLabel)
)
test_loss, test_acc = model.evaluate(set, label)
setSize = 10000
train_acc = 1
best = 0
while test_acc < .9:
del set
del label
set,label = G.generateNumpySet(setSize,G.letters, G.fonts) #Create new, smaller set to validate on
set = set/255.0
#G.shuffle(set,label)
model.fit(set, label,
epochs=100,
callbacks=[logger, earlyStop],
validation_data=(ValidationSet,ValidationLabel)
)
test_loss, test_acc = model.evaluate(ValidationSet, ValidationLabel, callbacks=[logger])
if test_acc > best:
print('Saved model to disc')
model.save('Models/'+MODEL_NAME)
best = test_acc
def plot_image(i, predictions_array, true_label, img):
predictions_array, true_label, img = predictions_array[i], true_label[i], img[i]
plt.grid(False)
plt.xticks([])
plt.yticks([])
plt.imshow(img, cmap=plt.cm.binary)
predicted_label = np.argmax(predictions_array)
if predicted_label == true_label:
color = 'blue'
else:
color = 'red'
plt.xlabel("{} {:2.0f}% ({})".format(G.letters[predicted_label],
100*np.max(predictions_array),
G.letters[true_label]),
color=color)
def plot_value_array(i, predictions_array, true_label):
predictions_array, true_label = predictions_array[i], true_label[i]
plt.grid(False)
plt.xticks([])
plt.yticks([])
thisplot = plt.bar(range(len(G.letters)), predictions_array, color="#777777")
plt.ylim([0, 1])
predicted_label = np.argmax(predictions_array)
thisplot[predicted_label].set_color('red')
thisplot[true_label].set_color('blue')
predictions = model.predict(ValidationSet)
num_rows = 5
num_cols = 3
num_images = num_rows*num_cols
plt.figure(figsize=(2*2*num_cols, 2*num_rows))
for i in range(num_images):
plt.subplot(num_rows, 2*num_cols, 2*i+1)
plot_image(i, predictions, ValidationLabel, ValidationSet)
plt.subplot(num_rows, 2*num_cols, 2*i+2)
plot_value_array(i, predictions, ValidationLabel)
plt.show()