Fall-Detection-System/deep_models.py at master · mojtabaSefidi/Fall-Detection-System · 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
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
from tensorflow import keras
import tensorflow as tf
import numpy as np
import time
from sklearn.metrics import classification_report, confusion_matrix

from utils import (CLASS_LABELS, CLASS_NAMES, plot_auc_curve,
                   plot_confusion_matrix, plot_history, plot_metrics,
                   plot_precision_recall_curve, plot_roc_curve)

class Train_Evaluate_Deep():

  def __init__(self,
               predictions=None,
               results=None,
               probabilities=None,
               performance=None):

      self.predictions = {} if predictions is None else predictions
      self.results = {} if results is None else results
      self.probabilities = {} if probabilities is None else probabilities
      self.performance = {} if performance is None else performance

  def build_cnn(self, input_size, units=128, drop_rate=0.25, filter=32, kernel_size=(1*9), output_size=3):
    input = keras.layers.Input((input_size))
    x = keras.layers.Conv1D(filters=filter//2, kernel_size=kernel_size, padding='same', activation='relu', name="conv1")(input)
    x = keras.layers.Conv1D(filters=filter, kernel_size=kernel_size, padding='same', activation='relu', name="conv2")(x)
    x = keras.layers.Conv1D(filters=filter*2, kernel_size=kernel_size, padding='same', activation='relu', name="conv3")(x)
    x = tf.keras.layers.Flatten()(x)
    classifier = keras.layers.Dense(units*4, activation='relu')(x)
    classifier = keras.layers.Dropout(drop_rate)(classifier)
    classifier = keras.layers.Dense(units, activation='relu')(classifier)
    classifier = keras.layers.Dropout(drop_rate)(classifier)
    classifier = keras.layers.Dense(units//2, activation='relu')(classifier)
    classifier = keras.layers.Dropout(drop_rate)(classifier)
    output = keras.layers.Dense(output_size, activation='softmax')(classifier)
    model = keras.Model(inputs=input, outputs=output)
    return model

  def build_lstm(self, input_size, units=128, drop_rate=0.25, lstm_units=16, output_size=3):
    input = keras.layers.Input((input_size))
    x = keras.layers.LSTM(units=lstm_units//2, input_shape=input_size, return_sequences=True, name="lstm1")(input)
    x = keras.layers.LSTM(units=lstm_units, input_shape=input_size, return_sequences=True, name="lstm2")(x)
    x = keras.layers.LSTM(units=lstm_units*2, input_shape=input_size, return_sequences=True, name="lstm3")(x)
    x = tf.keras.layers.Flatten()(x)
    classifier = keras.layers.Dense(units*4, activation='relu')(x)
    classifier = keras.layers.Dropout(drop_rate)(classifier)
    classifier = keras.layers.Dense(units, activation='relu')(classifier)
    classifier = keras.layers.Dropout(drop_rate)(classifier)
    classifier = keras.layers.Dense(units//2, activation='relu')(classifier)
    classifier = keras.layers.Dropout(drop_rate)(classifier)
    output = keras.layers.Dense(output_size, activation='softmax')(classifier)
    model = keras.Model(inputs=input, outputs=output)
    return model

  def build_mlp(self, input_size, hidden_layer_size=128, output_size=3, drop_rate=0.25):
    input = keras.layers.Input((input_size))
    x = keras.layers.Dense(hidden_layer_size//4, activation='relu')(input)
    x = keras.layers.Dense(hidden_layer_size//2, activation='relu')(x)
    x = keras.layers.Dense(hidden_layer_size, activation='relu')(x)
    x = tf.keras.layers.Flatten()(x)
    classifier = keras.layers.Dense(hidden_layer_size*4, activation='relu')(x)
    classifier = keras.layers.Dropout(drop_rate)(classifier)
    classifier = keras.layers.Dense(hidden_layer_size, activation='relu')(classifier)
    classifier = keras.layers.Dropout(drop_rate)(classifier)
    classifier = keras.layers.Dense(hidden_layer_size//2, activation='relu')(classifier)
    classifier = keras.layers.Dropout(drop_rate)(classifier)
    output = keras.layers.Dense(output_size, activation='softmax')(classifier)
    model = keras.Model(inputs=input, outputs=output)
    return model

  def build_AE(self, input_size, latent_dim=128, filter=64, kernel_size=(1*9)):
    input = keras.layers.Input((input_size))
    encoder = keras.layers.Conv1D(filters=filter//2, kernel_size=kernel_size, padding='same', activation='relu', name="conv1")(input)
    encoder = keras.layers.Conv1D(filters=filter, kernel_size=kernel_size, padding='same', activation='relu', name="conv2")(encoder)
    encoder = tf.keras.layers.Flatten()(encoder)
    encoder = keras.layers.Dense(latent_dim*4, activation='relu')(encoder)
    encoder = keras.layers.Dense(latent_dim*2, activation='relu')(encoder)

    latent = keras.layers.Dense(latent_dim, activation='relu')(encoder)

    decoder = keras.layers.Dense(latent_dim*2, activation='relu')(latent)
    decoder = keras.layers.Dense(latent_dim*4, activation='relu')(decoder)
    decoder = keras.layers.Dense(filter*input_size[0], activation='relu')(decoder)
    decoder = keras.layers.Reshape((input_size[0], filter))(decoder)
    decoder = keras.layers.Conv1DTranspose(filters=filter, kernel_size=kernel_size, padding='same', activation='relu')(decoder)
    decoder = keras.layers.Conv1DTranspose(filters=filter//2, kernel_size=kernel_size, padding='same', activation='relu')(decoder)

    output = keras.layers.Conv1DTranspose(filters=input_size[-1], kernel_size=kernel_size, padding='same', activation='relu')(decoder)
    model = keras.Model(inputs=input, outputs=output)
    return model

  def train_deep_model(self,
                       model,
                       X_train,
                       y_train,
                       metrics,
                       loss_function,
                       optimizer,
                       callbacks,
                       class_weight=None,
                       epochs=100,
                       batch_size=128,
                       validation_split=0.2,
                       validation_data=None,
                       verbose=2):

    print('----------------------------------')
    model.summary()
    print('----------------------------------')

    model.compile(optimizer=optimizer,
                  loss=loss_function,
                  metrics=metrics)

    fit_arguments = {
        'batch_size': batch_size,
        'epochs': epochs,
        'shuffle': True,
        'callbacks': callbacks,
        'verbose': verbose,
    }
    if validation_data is None:
      fit_arguments['validation_split'] = validation_split
    else:
      fit_arguments['validation_data'] = validation_data
    if class_weight is not None:
      fit_arguments['class_weight'] = class_weight
    return model.fit(X_train, y_train, **fit_arguments)

  def evaluate(self, model, X_test, y_test, batch_size=128, title='', model_name='',
               plot=True, training_seconds=None):
    model.predict(X_test[:min(batch_size, len(X_test))], batch_size=batch_size,
                  verbose=0)
    inference_started = time.perf_counter()
    prediction_scores = model.predict(X_test, batch_size=batch_size, verbose=0)
    inference_seconds = time.perf_counter() - inference_started
    prediction = np.argmax(prediction_scores, axis=1)
    self.predictions[model_name] = prediction
    self.probabilities[model_name] = prediction_scores
    if plot:
      print()
      print('1) Plot ROC Curve...')
      print()
      plot_roc_curve(y_test, prediction_scores, title='ROC Curve of {model_name} Model'.format(model_name=model_name), model_name=model_name, file_name=None)
      print()
      print('2) Plot AUC Curve...')
      print()
      plot_auc_curve(y_test, prediction_scores, title='AUC Curve of {model_name} Model'.format(model_name=model_name), model_name=model_name, file_name = None)
      print()
      print('3) Plot Percision_Recall Curve......')
      print()
      plot_precision_recall_curve(y_test, prediction_scores, title='Percision_Recall Curve of {model_name} Model'.format(model_name=model_name), model_name=model_name, file_name=None)
    report = classification_report(
        y_test, prediction, labels=CLASS_LABELS, target_names=CLASS_NAMES,
        output_dict=True, zero_division=0)
    self.results[model_name] = report
    self.performance[model_name] = {
        'training_seconds': training_seconds,
        'inference_seconds': inference_seconds,
        'inference_ms_per_window': 1000 * inference_seconds / max(len(X_test), 1),
        'model_size_bytes': int(sum(np.asarray(weight).nbytes for weight in model.get_weights())),
        'parameter_count': int(model.count_params()),
        'probability_classes': [0, 1, 2],
    }
    if plot:
      print()
      matrix = confusion_matrix(y_test, prediction, labels=CLASS_LABELS)
      plot_confusion_matrix(matrix, title=title, class_names=CLASS_NAMES)
      print()
    return prediction

  def plot_learning_curves(self, history):
    print('1) Plot learning process based on different metrics...')
    print()
    plot_metrics(history)
    print()
    print('2) Plot learning curve...')
    print()
    return plot_history(history)