tensorflow-facialexpr-recognition/train_landmarks.py at master · DataScienceUP/tensorflow-facialexpr-recognition · 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
import os
import glob

import dlib

import numpy as np
import tensorflow as tf

from lib.data_loader import _CNN_Data_Loader
from lib.config import ConfigReader, TrainNetConfig, DataConfig
from lib.CNNS.land_marks import _facenet

#NUM_PARALLEL_EXEC_UNITS = 4


# p = '/Users/miguelangelalbaacosta/Downloads/shape_predictor_68_face_landmarks.dat'
p = '/home/miguel_alba/data/shape_predictor_68_face_landmarks.dat'
detector = dlib.get_frontal_face_detector()
predictor = dlib.shape_predictor(p)
face_rects = [dlib.rectangle(left=1, top=1, right=47, bottom=47)]


def get_landmarks(image, rects):
    # this function have been copied from http://bit.ly/2cj7Fpq
    if len(rects) > 1:
        raise BaseException("TooManyFaces")
    if len(rects) == 0:
        raise BaseException("NoFaces")
    return np.matrix([[p.x, p.y] for p in predictor(image, rects[0]).parts()])


def create_batch_landmarks(crops):
    landmarks = []
    for ex in crops:
        lands = get_landmarks(ex[:, :, 0], face_rects)
        landmarks.append(lands)
    batch = np.array(landmarks).astype(np.float32)
    return batch


def train(conf_path):
    conf_path = conf_path
    config_reader = ConfigReader(conf_path)
    train_config = TrainNetConfig(config_reader.get_train_config())
    data_config = DataConfig(config_reader.get_train_config())

    out_dir = os.path.join(train_config.checkpoint_dir, 'models', train_config.name)
    train_log_dir = '{}/logs/train/'.format(out_dir)
    test_log_dir = '{}/logs/test/'.format(out_dir)

    if not os.path.exists(train_log_dir):
        os.makedirs(train_log_dir)
    if not os.path.exists(test_log_dir):
        os.makedirs(test_log_dir)

    net = _facenet(train_config)

    with tf.name_scope('input'):
        train_loader = _CNN_Data_Loader(data_config, name='train', training_mode=True, shuffle=True)
        train_image_batch, train_label_batch = train_loader._generate_batch()
        test_loader = _CNN_Data_Loader(data_config, name='test', training_mode=False, shuffle=False)  # default false
        test_image_batch, test_label_batch = test_loader._generate_batch()

    loss, accuracy = net.batch_model()
    train_op = net.optimize(loss)

    summary_op = tf.summary.merge_all()
    saver = tf.train.Saver(tf.global_variables())
    init = tf.global_variables_initializer()

    sess = tf.Session()
    sess.run(init)

    train_summary_writer = tf.summary.FileWriter(train_log_dir, sess.graph)
    val_summary_writer = tf.summary.FileWriter(test_log_dir, sess.graph)

    try:
        for step in np.arange(train_config.max_step):
            train_image, train_label = sess.run([train_image_batch, train_label_batch])
            train_land_marks = create_batch_landmarks(train_image)
            assert train_label.shape[1] == data_config.n_classes
            _, train_loss, train_acc = sess.run([train_op, loss, accuracy],
                                                feed_dict={net.x: train_image, net.y: train_label,
                                                           net.landmark: train_land_marks})
            if step % 50 == 0 or step + 1 == train_config.max_step:
                print('===TRAIN===: Step: %d, loss: %.4f, accuracy: %.4f%%' % (step, train_loss, train_acc))
                summary_str = sess.run(summary_op, feed_dict={net.x: train_image, net.y: train_label,
                                                              net.landmark: train_land_marks})
                train_summary_writer.add_summary(summary_str, step)
            if step % 150 == 0 or step + 1 == train_config.max_step:
                val_image, val_label = sess.run([test_image_batch, test_label_batch])
                val_land_marks = create_batch_landmarks(val_image)
                plot_images = tf.summary.image('val_images_{}'.format(step % 200), val_image, 10)
                val_loss, val_acc, plot_summary = sess.run([loss, accuracy, plot_images],
                                                           feed_dict={net.x: val_image, net.y: val_label,
                                                                      net.landmark: val_land_marks})
                print('====VAL====: Step %d, val loss = %.4f, val accuracy = %.4f%%' % (step, val_loss, val_acc))
                summary_str = sess.run(summary_op, feed_dict={net.x: val_image, net.y: val_label,
                                                              net.landmark: val_land_marks})
                val_summary_writer.add_summary(summary_str, step)
                val_summary_writer.add_summary(plot_summary, step)
            if step % 2000 == 0 or step + 1 == train_config.max_step:
                checkpoint_path = os.path.join(train_log_dir, 'model.ckpt')
                saver.save(sess, checkpoint_path, global_step=step)
    except tf.errors.OutOfRangeError:
        print('===INFO====: Training completed, reaching the maximum number of steps')
    sess.close()
    return None


def _train_201909_():
    conf_path = '/Volumes/SSD_ML/facialexpr/lib/experiments/experiment_2.yml'
    train(conf_path)
    return None


def train_cloud():
    conf_path = '/home/miguel_alba/facialexpr/lib/experiments/experiment_cloud.yml'
    train(conf_path)
    return None


def main():
    #_train_201909_()
    train_cloud()


if __name__ == '__main__':
    main()