DeepSHM/modelling.py at master · vewald/DeepSHM · GitHub

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
import os
os.environ['CUDA_VISIBLE_DEVICES'] = '0'

import keras
import tensorflow as tf
from keras.preprocessing.image import ImageDataGenerator
from keras.models import Sequential
from keras.layers import Conv2D, MaxPooling2D, Dropout, Flatten, Dense, LeakyReLU, GlobalAveragePooling2D
from keras import backend as K
from matplotlib import pyplot as plt
from IPython.display import clear_output

img_width, img_height = 3201, 193
# Variable window length for multiple scenarios:
    #101,7 #151,10 #201,13 #251,16 #301,19 #401,25 #501,31 #601,37 #801,49
    #1001,61 #1201,73 #1601,97 #2001,121 #3201,193
train_data_dir = 'data/multiple_scene1/3200/sensor1/training'
validation_data_dir = 'data/multiple_scene1/3200/sensor1/validation'

num_classes = 6
nb_train_samples = num_classes*70
nb_validation_samples = num_classes*20
epochs = 50
batch_size = 32

if K.image_data_format() == 'channels_first':
    input_shape = (3, img_width, img_height)
else:
    input_shape = (img_width, img_height, 3)

class PlotLearning(keras.callbacks.Callback):
    def on_train_begin(self, logs={}):
        self.i = 0
        self.x = []
        self.losses = []
        self.val_losses = []
        self.acc = []
        self.val_acc = []
        self.fig = plt.figure()

        self.logs = []

    def on_epoch_end(self, epoch, logs={}):

        self.logs.append(logs)
        self.x.append(self.i)
        self.losses.append(logs.get('loss'))
        self.val_losses.append(logs.get('val_loss'))
        self.acc.append(logs.get('categorical_accuracy'))
        self.val_acc.append(logs.get('val_categorical_accuracy'))
        self.i += 1

        clear_output(wait=True)
        color1 = 'tab:red'
        color2 = 'tab:blue'
        fig, ax1 = plt.subplots(figsize=(10, 6))
        ax1.set_xlabel('Epoch',size=24)
        ax1.set_ylabel('Loss',color=color1,size=24)
        ax1.plot(self.x, self.losses, label="tr_loss",color=color1,linestyle='dashed')
        ax1.plot(self.x, self.val_losses, label="val_loss",color=color1)
        ax1.tick_params(axis='x', labelsize = 16)
        ax1.tick_params(axis='y', labelcolor=color1, labelsize = 14)
        ax1.legend(loc='center right',fontsize=16,bbox_to_anchor=(0.4, 1.1),ncol = 2)
        ax1.set_ylim([0, 1.05])
        ax2 = ax1.twinx()
        ax2.set_ylabel('Accuracy',color=color2,size=24)
        ax2.plot(self.x, self.acc, label="tr_accuracy",color=color2,linestyle='dashed')
        ax2.plot(self.x, self.val_acc, label="val_accuracy",color=color2)
        ax2.tick_params(axis='y', labelcolor=color2, labelsize = 16)
        ax2.legend(loc='center right',fontsize=16, bbox_to_anchor=(1.1, 1.1),ncol = 2)
        ax2.set_ylim([0, 3])
        fig.tight_layout()

        plt.show();

#plot_losses = PlotLearning()

x_filter = 3
y_filter = 3
stride = 1
x_pool = 2
y_pool = 2
do_rate = 0.50

model = Sequential()

model.add(Conv2D(8, (x_filter, y_filter), strides = (stride, stride), activation='relu', padding='same',kernel_initializer='random_uniform', input_shape=input_shape))
model.add(MaxPooling2D((x_pool, y_pool), padding='same'))
model.add(Dropout(do_rate))
model.add(Conv2D(16, (x_filter, y_filter), strides = (stride, stride), activation='relu', padding='same'))
model.add(MaxPooling2D((x_pool, y_pool), padding='same'))
model.add(Dropout(do_rate))
model.add(Conv2D(32, (x_filter, y_filter), strides = (stride, stride), activation='relu', padding='same'))
model.add(MaxPooling2D((x_pool, y_pool), padding='same'))
model.add(Dropout(do_rate))
model.add(Flatten())
#model.add(GlobalAveragePooling2D())
model.add(Dense(128, activation='relu'))
model.add(Dropout(do_rate))
model.add(Dense(16, activation='relu'))
model.add(Dropout(do_rate))
model.add(Dense(num_classes, activation='softmax'))

model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['categorical_accuracy'])
model.summary()

train_datagen = ImageDataGenerator(rescale = 1. / 255)#, width_shift_range = 0.1)
test_datagen = ImageDataGenerator(rescale = 1. / 255)#, width_shift_range = 0.1)

train_generator = train_datagen.flow_from_directory(
    train_data_dir, target_size=(img_width, img_height),
    batch_size=batch_size, color_mode='rgb', class_mode='categorical')

validation_generator = test_datagen.flow_from_directory(
    validation_data_dir, target_size=(img_width, img_height),
    batch_size=batch_size, color_mode='rgb', class_mode='categorical')

model.fit_generator(
    train_generator,
    steps_per_epoch=nb_train_samples // batch_size,
    epochs=epochs,
    #callbacks=[plot_losses],
    validation_data=validation_generator,
    validation_steps=nb_validation_samples // batch_size)

model_json = model.to_json()
with open("model.json", "w") as json_file:
    json_file.write(model_json)
# serialize weights to HDF5
#model.save("F://Spyder/Perception_Modelling/src")
#print("Saved model to disk")