Image-Captioning/train_Model.py at master · sainimohit23/Image-Captioning · 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
import keras
import keras.backend as K
import numpy as np
import pickle
from keras.preprocessing.sequence import pad_sequences
import tensorflow as tf
from preprocess_utils import preprocess, preprocess_input
import matplotlib.pyplot as plt


def load_preprocess(filename):
    with open(filename, mode='rb') as in_file:
        return pickle.load(in_file)

train_inputs, test_inputs, vocab_to_int, int_to_vocab = load_preprocess('preprocess.p')
transfer_values_train = load_preprocess('encoded_images_vgg16.p')
transfer_values_train = np.array(transfer_values_train)


transfer_values_test = load_preprocess('encoded_images_test_vgg16.p')
transfer_values_test = np.array(transfer_values_test)


def get_random_tokens(idx):
    result = []

    for i in idx:
        j = np.random.choice(len(train_inputs[i]))

        result.append(train_inputs[i][j])

    return result


def batch_generator(batch_size):
    while True:
        idx = np.random.randint(len(train_inputs), size= batch_size)
        transfer_values = transfer_values_train[idx]
        tokens = get_random_tokens(idx)

        token_lengths = [len(t) for t in tokens]
        max_len = np.max(token_lengths)

        tokens_padded = pad_sequences(tokens,
                                      maxlen= max_len,
                                      padding='post',
                                      truncating='post',
                                      value= vocab_to_int['<PAD>'])


        decoder_input_data = tokens_padded[:, 0:-1]
        decoder_output_data = tokens_padded[:, 1:]


        x_data =  {'decoder_input': decoder_input_data,
                   'transfer_values_input': transfer_values}

        y_data ={'decoder_output': decoder_output_data}

        yield(x_data, y_data)


state_size = 512
embedding_size = 128
transfer_values_size = transfer_values_train[0].shape[0]
num_words = len(int_to_vocab) + 1
batch_size = 1024


generator = batch_generator(batch_size=batch_size)


transfer_values_input = keras.layers.Input(shape=(transfer_values_size,), name='transfer_values_input')

decoder_transfer_map = keras.layers.Dense(state_size,
                                          activation='tanh',
                                          name='decoder_transfer_map')

decoder_input = keras.layers.Input(shape=(None, ), name='decoder_input')

decoder_embedding = keras.layers.embeddings.Embedding(input_dim=num_words,
                                                      output_dim=embedding_size,
                                                      name='decoder_embedding')

decoder_gru1 = keras.layers.GRU(state_size, name='decoder_gru1',
                                return_sequences=True)
decoder_gru2 = keras.layers.GRU(state_size, name='decoder_gru2',
                                return_sequences=True)
decoder_gru3 = keras.layers.GRU(state_size, name='decoder_gru3',
                                return_sequences=True)

decoder_dense = keras.layers.Dense(num_words,
                      activation='linear',
                      name='decoder_output')

def connect_decoder(transfer_values):

    initial_state = decoder_transfer_map(transfer_values)

    net = decoder_input
    net = decoder_embedding(net)

    # Connect all the GRU layers.
    net = decoder_gru1(net, initial_state=initial_state)
    net = decoder_gru2(net, initial_state=initial_state)

    net = keras.layers.Dropout(0.5)(net)
    net = decoder_gru3(net, initial_state=initial_state)

    net = keras.layers.Dropout(0.5)(net)
    decoder_output = decoder_dense(net)

    return decoder_output

decoder_output = connect_decoder(transfer_values=transfer_values_input)

decoder_model = keras.models.Model(inputs=[transfer_values_input, decoder_input],
                      outputs=[decoder_output])


def sparse_cross_entropy(y_true, y_pred):

    loss = tf.nn.sparse_softmax_cross_entropy_with_logits(labels=y_true,
                                                          logits=y_pred)

    loss_mean = tf.reduce_mean(loss)
    return loss_mean

optimizer = keras.optimizers.RMSprop(lr=1e-3)
decoder_target = tf.placeholder(dtype='int32', shape=(None, None))
decoder_model.compile(optimizer=optimizer,
                      loss=sparse_cross_entropy,
                      target_tensors=[decoder_target])

try:
    decoder_model.load_weights('Captioning.h5')
except Exception:
    print('Trained Model not found, retraining model')

decoder_model.fit_generator(generator=generator,
                            steps_per_epoch=30,
                            epochs=20)