-
Notifications
You must be signed in to change notification settings - Fork 246
/
training_fracture.py
190 lines (164 loc) · 7.13 KB
/
training_fracture.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
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
import numpy as np
import pandas as pd
import os.path
import matplotlib.pyplot as plt
from sklearn.model_selection import train_test_split
import tensorflow as tf
from tensorflow.keras.optimizers import Adam
# load images to build and train the model
# .... / img1.jpg
# test Hand patient0000 positive -- img2.png
# / / \ .....
# Dataset - Elbow ------ patient0001
# \ train \ / img1.png
# Shoulder patient0002 negative -- img2.jpg
# .... \
#
def load_path(path, part):
"""
load X-ray dataset
"""
dataset = []
for folder in os.listdir(path):
folder = path + '/' + str(folder)
if os.path.isdir(folder):
for body in os.listdir(folder):
if body == part:
body_part = body
path_p = folder + '/' + str(body)
for id_p in os.listdir(path_p):
patient_id = id_p
path_id = path_p + '/' + str(id_p)
for lab in os.listdir(path_id):
if lab.split('_')[-1] == 'positive':
label = 'fractured'
elif lab.split('_')[-1] == 'negative':
label = 'normal'
path_l = path_id + '/' + str(lab)
for img in os.listdir(path_l):
img_path = path_l + '/' + str(img)
dataset.append(
{
'body_part': body_part,
'patient_id': patient_id,
'label': label,
'image_path': img_path
}
)
return dataset
# this function get part and know what kind of part to train, save model and save plots
def trainPart(part):
# categories = ['fractured', 'normal']
THIS_FOLDER = os.path.dirname(os.path.abspath(__file__))
image_dir = THIS_FOLDER + '/Dataset/'
data = load_path(image_dir, part)
labels = []
filepaths = []
# add labels for dataframe for each category 0-fractured, 1- normal
for row in data:
labels.append(row['label'])
filepaths.append(row['image_path'])
filepaths = pd.Series(filepaths, name='Filepath').astype(str)
labels = pd.Series(labels, name='Label')
images = pd.concat([filepaths, labels], axis=1)
# split all dataset 10% test, 90% train (after that the 90% train will split to 20% validation and 80% train
train_df, test_df = train_test_split(images, train_size=0.9, shuffle=True, random_state=1)
# each generator to process and convert the filepaths into image arrays,
# and the labels into one-hot encoded labels.
# The resulting generators can then be used to train and evaluate a deep learning model.
# now we have 10% test, 72% training and 18% validation
train_generator = tf.keras.preprocessing.image.ImageDataGenerator(horizontal_flip=True,
preprocessing_function=tf.keras.applications.resnet50.preprocess_input,
validation_split=0.2)
# use ResNet50 architecture
test_generator = tf.keras.preprocessing.image.ImageDataGenerator(
preprocessing_function=tf.keras.applications.resnet50.preprocess_input)
train_images = train_generator.flow_from_dataframe(
dataframe=train_df,
x_col='Filepath',
y_col='Label',
target_size=(224, 224),
color_mode='rgb',
class_mode='categorical',
batch_size=64,
shuffle=True,
seed=42,
subset='training'
)
val_images = train_generator.flow_from_dataframe(
dataframe=train_df,
x_col='Filepath',
y_col='Label',
target_size=(224, 224),
color_mode='rgb',
class_mode='categorical',
batch_size=64,
shuffle=True,
seed=42,
subset='validation'
)
test_images = test_generator.flow_from_dataframe(
dataframe=test_df,
x_col='Filepath',
y_col='Label',
target_size=(224, 224),
color_mode='rgb',
class_mode='categorical',
batch_size=32,
shuffle=False
)
# we use rgb 3 channels and 224x224 pixels images, use feature extracting , and average pooling
pretrained_model = tf.keras.applications.resnet50.ResNet50(
input_shape=(224, 224, 3),
include_top=False,
weights='imagenet',
pooling='avg')
# for faster performance
pretrained_model.trainable = False
inputs = pretrained_model.input
x = tf.keras.layers.Dense(128, activation='relu')(pretrained_model.output)
x = tf.keras.layers.Dense(50, activation='relu')(x)
# outputs Dense '2' because of 2 classes, fratured and normal
outputs = tf.keras.layers.Dense(2, activation='softmax')(x)
model = tf.keras.Model(inputs, outputs)
# print(model.summary())
print("-------Training " + part + "-------")
# Adam optimizer with low learning rate for better accuracy
model.compile(optimizer=Adam(learning_rate=0.0001), loss='categorical_crossentropy', metrics=['accuracy'])
# early stop when our model is over fit or vanishing gradient, with restore best values
callbacks = tf.keras.callbacks.EarlyStopping(monitor='val_loss', patience=3, restore_best_weights=True)
history = model.fit(train_images, validation_data=val_images, epochs=25, callbacks=[callbacks])
# save model to this path
model.save(THIS_FOLDER + "/weights/ResNet50_" + part + "_frac.h5")
results = model.evaluate(test_images, verbose=0)
print(part + " Results:")
print(results)
print(f"Test Accuracy: {np.round(results[1] * 100, 2)}%")
# create plots for accuracy and save it
plt.plot(history.history['accuracy'])
plt.plot(history.history['val_accuracy'])
plt.title('model accuracy')
plt.ylabel('accuracy')
plt.xlabel('epoch')
plt.legend(['train', 'test'], loc='upper left')
# plt.show()
figAcc = plt.gcf()
my_file = os.path.join(THIS_FOLDER, "./plots/FractureDetection/" + part + "/_Accuracy.jpeg")
figAcc.savefig(my_file)
plt.clf()
# create plots for loss and save it
plt.plot(history.history['loss'])
plt.plot(history.history['val_loss'])
plt.title('model loss')
plt.ylabel('loss')
plt.xlabel('epoch')
plt.legend(['train', 'test'], loc='upper left')
# plt.show()
figAcc = plt.gcf()
my_file = os.path.join(THIS_FOLDER, "./plots/FractureDetection/" + part + "/_Loss.jpeg")
figAcc.savefig(my_file)
plt.clf()
# run the function and create model for each parts in the array
categories_parts = ["Elbow", "Hand", "Shoulder"]
for category in categories_parts:
trainPart(category)