agender_model.py

from keras import applications
from keras.models import Model
from keras.layers import Dense
from keras.optimizer_v1 import SGD, Adam


def get_model(cfg):
    base_model = getattr(applications, cfg.model.model_name)(
        include_top=False,
        input_shape=(cfg.model.img_size, cfg.model.img_size, 3),
        pooling="avg"
    )
    features = base_model.output
    pred_gender = Dense(units=2, activation="softmax", name="pred_gender")(features)
    pred_age = Dense(units=101, activation="softmax", name="pred_age")(features)
    model = Model(inputs=base_model.input, outputs=[pred_gender, pred_age])
    return model


def get_optimizer(cfg):
    if cfg.train.optimizer_name == "sgd":
        return SGD(lr=cfg.train.lr, momentum=0.9, nesterov=True)
    elif cfg.train.optimizer_name == "adam":
        return Adam(lr=cfg.train.lr)
    else:
        raise ValueError("optimizer name should be 'sgd' or 'adam'")


def get_scheduler(cfg):
    class Schedule:
        def __init__(self, nb_epochs, initial_lr):
            self.epochs = nb_epochs
            self.initial_lr = initial_lr

        def __call__(self, epoch_idx):
            if epoch_idx < self.epochs * 0.25:
                return self.initial_lr
            elif epoch_idx < self.epochs * 0.50:
                return self.initial_lr * 0.2
            elif epoch_idx < self.epochs * 0.75:
                return self.initial_lr * 0.04
            return self.initial_lr * 0.008

    return Schedule(cfg.train.epochs, cfg.train.lr)