05_RealAge_v3

from tensorflow.keras.applications.resnet import ResNet50
from tensorflow.keras.layers import Dense, GlobalAveragePooling2D
from tensorflow.keras.models import Sequential
from tensorflow.keras.preprocessing.image import ImageDataGenerator
from tensorflow.keras.optimizers import Adam
import pandas as pd

train_datagen = ImageDataGenerator(rescale=1./255,
                                   horizontal_flip=True,
                                   zoom_range=0.2,
                                   )

test_datagen = ImageDataGenerator(rescale=1./255)

def load_train(path):
    labels = pd.read_csv(path + 'labels.csv')
    train_datagen_flow = train_datagen.flow_from_dataframe(
    dataframe=labels,
    directory=path + '/final_files',
    x_col='file_name',
    y_col='real_age',
    target_size=(224, 224),
    batch_size=64,
    class_mode='raw',
    subset='training',
    seed=42)
    return train_datagen_flow

def load_test(path):
    labels = pd.read_csv(path + 'labels.csv')
    test_datagen_flow = test_datagen.flow_from_dataframe(
    dataframe=labels,
    directory=path + '/final_files',
    x_col='file_name',
    y_col='real_age',
    target_size=(224, 224),
    batch_size=32,
    class_mode='raw',
    subset='validation',
    seed=42)
    return test_datagen_flow

def create_model(input_shape):
    optimizer = Adam(learning_rate=0.0001)
    cnn = ResNet50(input_shape=input_shape, include_top=False,
        weights='/datasets/keras_models/resnet50_weights_tf_dim_ordering_tf_kernels_notop.h5',
        )

    model = Sequential()
    model.add(cnn)
    model.add(GlobalAveragePooling2D())
    model.add(Dense(1, activation="relu"))
    model.compile(loss="mean_absolute_error", optimizer=optimizer, metrics=["mae"])
    return model

def train_model(model, train_data, test_data, batch_size=None, epochs=10,
               steps_per_epoch=None, validation_steps=None):
    model.fit(train_data,
              validation_data=test_data,
              batch_size=batch_size,
              epochs=epochs,
              steps_per_epoch=steps_per_epoch,
              verbose=2, shuffle=True)
    return model