/
train.py
116 lines (83 loc) · 3.58 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
#__author__ = ritvikareddy2
#__date__ = 2018-12-09
import cv2
import itertools
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import keras
from keras.callbacks import ModelCheckpoint
from keras.layers import Activation, BatchNormalization, Conv2D, Dense, Dropout, Flatten, MaxPool2D, \
Reshape
from keras.models import Sequential
from keras.utils import np_utils
from sklearn.metrics import confusion_matrix
from data_utils import process_data, convert_probs_to_labels
def load_data():
df = pd.read_csv("fer2013.csv")
return df
def train_and_evaluate_model():
data = load_data()
x_train, y_train = process_data(data, 'Training')
x_validation, y_validation = process_data(data, 'PublicTest')
model = Sequential()
model.add(Conv2D(64, 3, data_format="channels_last", kernel_initializer="he_normal",
input_shape=(48, 48, 1), activation='relu'))
model.add(BatchNormalization())
model.add(Conv2D(64, 3, activation='relu'))
model.add(BatchNormalization())
model.add(MaxPool2D(pool_size=(2, 2), strides=2))
model.add(Dropout(0.6))
model.add(Conv2D(32, 3, activation='relu'))
model.add(BatchNormalization())
model.add(Conv2D(32, 3, activation='relu'))
model.add(BatchNormalization())
model.add(Conv2D(32, 3, activation='relu'))
model.add(BatchNormalization())
model.add(MaxPool2D(pool_size=(2, 2), strides=2))
model.add(Dropout(0.6))
model.add(Flatten())
model.add(Dense(128, activation='relu'))
model.add(BatchNormalization())
model.add(Dropout(0.6))
model.add(Dense(7, activation='softmax'))
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
print(model.summary())
# save best weights
checkpoint = ModelCheckpoint(filepath='cnn_basic_2', verbose=1, save_best_only=True)
# num epochs
epochs = 20
# run model
model_history = model.fit(x_train, y_train, epochs=epochs,
shuffle=True,
batch_size=100, validation_data=(x_validation, y_validation),
callbacks=[checkpoint], verbose=2)
# save model to json
model_json = model.to_json()
with open("cnn2.json", "w") as json_file:
json_file.write(model_json)
fig = plt.figure(figsize=(12, 4))
plt.subplot(1, 2, 1)
plt.suptitle('CNN on raw pixel data', fontsize=10)
plt.ylabel('Loss', fontsize=16)
plt.plot(model_history.history['loss'], color='b', label='Training Loss')
plt.plot(model_history.history['val_loss'], color='r', label='Validation Loss')
plt.legend(loc='upper right')
plt.subplot(1, 2, 2)
plt.ylabel('Accuracy', fontsize=16)
plt.plot(model_history.history['acc'], color='b', label='Training Accuracy')
plt.plot(model_history.history['val_acc'], color='r', label='Validation Accuracy')
plt.legend(loc='lower right')
fig.savefig("Loss_Accuracy_Comparison.png")
plt.show()
x_test, y_test = process_data(data, "PrivateTest")
y_predicted_probs = model.predict(x_test, verbose=1)
emotion_labels = {0: "Angry", 1: "Disgust", 2: "Fear", 3: "Happy", 4: "Sad", 5: "Surprise",
6: "Neutral"}
labels = ["Angry", "Disgust", "Fear", "Happy", "Sad", "Surprise", "Neutral"]
score = model.evaluate(x_test, y_test, verbose=1)
print("Metrics: ", model.metrics_names)
print("Loss on test data", score[0])
print("Test Accuracy", score[1])
y_predicted_labels = convert_probs_to_labels(y_predicted_probs, labels)
y_actual_labels = convert_probs_to_labels(y_test, labels)