File size: 6,841 Bytes
aca4b90
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
# -*- coding: utf-8 -*-
"""
Created on Thu Feb 29 15:05:30 2024

@author: Dhrumit Patel
"""

"""
Dataset: UTKFace
https://www.kaggle.com/datasets/jangedoo/utkface-new?resource=download
"""

import pandas as pd
import numpy as np
import tensorflow as tf
import os
import matplotlib.pyplot as plt
import cv2
from keras.models import Sequential, Model, load_model
from keras.layers import Conv2D, MaxPool2D, Dense, Dropout, BatchNormalization, Flatten, Input
from sklearn.model_selection import train_test_split

path = 'data/UTKFace'

images = []
age = []
gender = []

for img in os.listdir(path):
    ages = img.split("_")[0]
    genders = img.split("_")[1]
    img = cv2.imread(str(path) + "/" + str(img))
    img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
    images.append(np.array(img))
    age.append(np.array(ages))
    gender.append(np.array(genders))

images = np.array(images)

images = images / 255.0
age = np.array(age, dtype=np.int64)
gender = np.array(gender, dtype=np.int64)

# np.save('preprocessed_data/images.npy', images)
# np.save('preprocessed_data/age.npy', age)
# np.save('preprocessed_data/gender.npy', gender)

# images = np.load('preprocessed_data/images.npy')
# age = np.load('preprocessed_data/age.npy')
# gender = np.load('preprocessed_data/gender.npy')

X_train_age, X_test_age, y_train_age, y_test_age = train_test_split(images, age, random_state=42)
X_train_gender, X_test_gender, y_train_gender, y_test_gender = train_test_split(images, gender, random_state=42)

# Define the model - For Age
age_model = Sequential()
age_model.add(Conv2D(128, kernel_size=3, activation='relu', input_shape=(200,200,3)))
age_model.add(MaxPool2D(pool_size=3, strides=2))

age_model.add(Conv2D(128, kernel_size=3, activation='relu'))
age_model.add(MaxPool2D(pool_size=3, strides=2))
              
age_model.add(Conv2D(256, kernel_size=3, activation='relu'))
age_model.add(MaxPool2D(pool_size=3, strides=2))

age_model.add(Conv2D(512, kernel_size=3, activation='relu'))
age_model.add(MaxPool2D(pool_size=3, strides=2))

age_model.add(Flatten())
age_model.add(Dropout(0.2))
age_model.add(Dense(512, activation='relu'))

age_model.add(Dense(1, activation='linear', name='age'))
              
age_model.compile(optimizer='adam', loss='mse', metrics=['mae'])

age_model.summary()

# TODO: Fit for 10 epochs
with tf.device('/CPU:0'):
    history_age = age_model.fit(X_train_age, y_train_age,
                                epochs=10,
                                validation_data=(X_test_age, y_test_age))  

age_model.save('models/age_model_10epochs.h5')

# Define the model - For Gender
gender_model = Sequential()

gender_model.add(Conv2D(36, kernel_size=3, activation='relu', input_shape=(200,200,3)))

gender_model.add(MaxPool2D(pool_size=3, strides=2))
gender_model.add(Conv2D(64, kernel_size=3, activation='relu'))
gender_model.add(MaxPool2D(pool_size=3, strides=2))

gender_model.add(Conv2D(128, kernel_size=3, activation='relu'))
gender_model.add(MaxPool2D(pool_size=3, strides=2))

gender_model.add(Conv2D(256, kernel_size=3, activation='relu'))
gender_model.add(MaxPool2D(pool_size=3, strides=2))

gender_model.add(Conv2D(512, kernel_size=3, activation='relu'))
gender_model.add(MaxPool2D(pool_size=3, strides=2))

gender_model.add(Flatten())
gender_model.add(Dropout(0.2))
gender_model.add(Dense(512, activation='relu'))
gender_model.add(Dense(1, activation='sigmoid', name='gender'))

gender_model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])

gender_model.summary()

# TODO: Fit for 10 epochs
with tf.device('/CPU:0'):
    history_gender = gender_model.fit(X_train_gender, y_train_gender,
                                      epochs=10,
                                      validation_data=(X_test_gender, y_test_gender))

gender_model.save('models/gender_model_10epochs.h5')

# Plot the training and validation accuracy and loss at each epoch - For Age model
loss = history_age.history['loss']
val_loss = history_age.history['val_loss']
epochs = range(1, len(loss) + 1)
plt.plot(epochs, loss, 'y', label='Training loss')
plt.plot(epochs, val_loss, 'r', label='Validation loss')
plt.title('Training and validation loss for Age model')
plt.xlabel('Epochs')
plt.ylabel('Loss')
plt.legend()
plt.show()

acc = history_age.history['accuracy']
val_acc = history_age.history['val_accuracy']
plt.plot(epochs, acc, 'y', label='Training Accuracy')
plt.plot(epochs, val_acc, 'r', label='Validation Accuracy')
plt.title('Training and validation accuracy for Age model')
plt.xlabel('Epochs')
plt.ylabel('Accuracy')
plt.legend()
plt.show()

# Plot the training and validation accuracy and loss at each epoch - For Gender model
loss = history_gender.history['loss']
val_loss = history_gender.history['val_loss']
epochs = range(1, len(loss) + 1)
plt.plot(epochs, loss, 'y', label='Training loss')
plt.plot(epochs, val_loss, 'r', label='Validation loss')
plt.title('Training and validation loss for Gender model')
plt.xlabel('Epochs')
plt.ylabel('Loss')
plt.legend()
plt.show()

acc = history_gender.history['accuracy']
val_acc = history_gender.history['val_accuracy']
plt.plot(epochs, acc, 'y', label='Training Accuracy')
plt.plot(epochs, val_acc, 'r', label='Validation Accuracy')
plt.title('Training and validation accuracy for Gender model')
plt.xlabel('Epochs')
plt.ylabel('Accuracy')
plt.legend()
plt.show()


# Test the model - Gender Model
my_model_gender = load_model('models/gender_model_3epochs.h5')
with tf.device('/CPU:0'):
    predictions = my_model_gender.predict(X_test_gender)
y_pred = (predictions >= 0.5).astype(int)[:, 0]

from sklearn.metrics import accuracy_score, confusion_matrix
import seaborn as sns

print(f"Accuracy: {accuracy_score(y_true=y_test_gender, y_pred=y_pred)}")

# Confusion matrix
cm = confusion_matrix(y_true=y_test_gender, y_pred=y_pred)
sns.heatmap(cm, annot=True, fmt='d')

# Test the model - Age Model

# Saving the models to huggingface
from transformers import TFAutoModel
import tensorflow as tf

# Load your TensorFlow model from the .h5 file
age_model = tf.keras.models.load_model('models/age_model_3epochs.h5')
gender_model = tf.keras.models.load_model('models/gender_model_3epochs.h5')

# Save the model weights
age_model.save_weights('age_model_weights.h5')
gender_model.save_weights('gender_model_weights.h5')

# Load the architecture of the Hugging Face model you want to use
# For example, if you're using BERT, you would use TFBertModel
hf_model_age = TFAutoModel.from_pretrained('bert-base-uncased')
hf_model_gender = TFAutoModel.from_pretrained('bert-base-uncased')

# Load the weights into the Hugging Face model
hf_model_age.load_weights('age_model_weights.h5')
hf_model_gender.load_weights('gender_model_weights.h5')

# Save the Hugging Face model
hf_model_age.save_pretrained('hf_age_model')
hf_model_gender.save_pretrained('hf_gender_model')