model.py

from tensorflow import keras
from tensorflow.keras.layers import (
    Dense,
    Conv2D,
    MaxPool2D,
    Flatten,
    Dropout,
    LocallyConnected2D,
)
import os
from pathlib import Path
import zipfile
import gdown


def get_input_shape():
    return (152, 152)


def load_FbDeepFace():
    # model = keras.Sequential()

    # model.add(Conv2D(filters=32, kernel_size=(11,11), activation='relu', name='C1', input_shape=(152,152,3)))
    # model.add(MaxPool2D(pool_size=(3,3), strides=2, padding='same', name='M2'))
    # model.add(Conv2D(filters=16, kernel_size=(9,9), activation='relu', name='C3'))

    # model.add(LocallyConnected2D(filters=16, kernel_size=(9,9), activation='relu', name='L4'))
    # model.add(LocallyConnected2D(filters=16, kernel_size=(7,7), strides=2, activation='relu', name='L5'))
    # model.add(LocallyConnected2D(filters=16, kernel_size=(5,5), activation='relu', name='L6'))

    # model.add(Flatten(name='F0'))

    # model.add(Dense(4096, activation='relu', name='F7'))
    # model.add(Dropout(rate=0.5, name='D0'))
    # model.add(Dense(8631, activation='softmax', name='F8'))

    # =====================================================================================

    # fmt: off
    deepface_inputs = keras.Input(shape=(152,152,3), name="In1")
    x = Conv2D(filters=32, kernel_size=(11,11), activation='relu', name='C1')(deepface_inputs)
    x = MaxPool2D(pool_size=(3,3), strides=2, padding='same', name='M2')(x)
    x = Conv2D(filters=16, kernel_size=(9,9), activation='relu', name='C3')(x)
    
    x = LocallyConnected2D(filters=16, kernel_size=(9,9), activation='relu', name='L4')(x)
    x = LocallyConnected2D(filters=16, kernel_size=(7,7), strides=2, activation='relu', name='L5')(x)
    x = LocallyConnected2D(filters=16, kernel_size=(5,5), activation='relu', name='L6')(x)
    
    flat = Flatten(name='F0')(x)
    
    embedding_layer = Dense(4096, activation='relu', name='F7')(flat)
    drop_1 = Dropout(rate=0.5, name='D0')(embedding_layer)
    deepface_outputs = Dense(8631, activation='softmax', name='F8')(drop_1)
    model = keras.Model(inputs=deepface_inputs, outputs=deepface_outputs)
    # fmt: on

    # Download the weights
    home = str(Path.home())
    model_path = home + "/.deepface/weights/VGGFace2_DeepFace_weights_val-0.9034.h5"

    if os.path.isfile(model_path) != True:
        print("DeepFace weights by will be downloaded ")
        os.makedirs(home + "/.deepface/weights")

        url = "https://github.com/swghosh/DeepFace/releases/download/weights-vggface2-2d-aligned/VGGFace2_DeepFace_weights_val-0.9034.h5.zip"

        output = model_path + ".zip"

        gdown.download(url, output, quiet=False)

        with zipfile.ZipFile(output, "r") as z:
            z.extractall(home + "/.deepface/weights")

    model.load_weights(model_path)

    fb_deepface_model = keras.Model(inputs=deepface_inputs, outputs=embedding_layer)

    return fb_deepface_model