{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"pygments_lexer":"ipython3","nbconvert_exporter":"python","version":"3.6.4","file_extension":".py","codemirror_mode":{"name":"ipython","version":3},"name":"python","mimetype":"text/x-python"}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"markdown","source":"# Wstępna konfiguracja\n\nNależy zaimportować odpowiednie biblioteki","metadata":{}},{"cell_type":"code","source":"import tensorflow as tf\nfrom tensorflow import keras\nfrom tensorflow.keras import layers\nimport tensorflow_addons as tfa\n\nfrom kaggle_datasets import KaggleDatasets\nimport matplotlib.pyplot as plt\nimport numpy as np\n\n","metadata":{"_uuid":"d629ff2d2480ee46fbb7e2d37f6b5fab8052498a","_cell_guid":"79c7e3d0-c299-4dcb-8224-4455121ee9b0","execution":{"iopub.status.busy":"2022-04-19T10:49:07.292917Z","iopub.execute_input":"2022-04-19T10:49:07.293257Z","iopub.status.idle":"2022-04-19T10:49:11.583009Z","shell.execute_reply.started":"2022-04-19T10:49:07.293223Z","shell.execute_reply":"2022-04-19T10:49:11.582139Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"TPU - Tensor Processing Unit, układ stworzony przez Google do optymalizacji uczenia maszynowego.\nJest to jeden z wielu rodzajów akceleratorów, czyli układów przyspieszających obliczenia.","metadata":{}},{"cell_type":"code","source":"try:\n    tpu = tf.distribute.cluster_resolver.TPUClusterResolver()\n    print('Device:', tpu.master())\n    tf.config.experimental_connect_to_cluster(tpu)\n    tf.tpu.experimental.initialize_tpu_system(tpu)\n    strategy = tf.distribute.experimental.TPUStrategy(tpu)\nexcept:\n    strategy = tf.distribute.get_strategy()\nprint('Number of replicas:', strategy.num_replicas_in_sync)\n\nAUTOTUNE = tf.data.experimental.AUTOTUNE\n    \nprint(tf.__version__)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Ładowanie danych\n\nDefiniowana jest ścieżka do zestawu danych.","metadata":{}},{"cell_type":"code","source":"GCS_PATH = KaggleDatasets().get_gcs_path()","metadata":{"execution":{"iopub.status.busy":"2022-04-19T10:49:39.944793Z","iopub.execute_input":"2022-04-19T10:49:39.945121Z","iopub.status.idle":"2022-04-19T10:49:40.312403Z","shell.execute_reply.started":"2022-04-19T10:49:39.945091Z","shell.execute_reply":"2022-04-19T10:49:40.31162Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"MONET_FILENAMES = tf.io.gfile.glob(str(GCS_PATH + '/monet_tfrec/*.tfrec'))\nprint('Monet Files:', len(MONET_FILENAMES))\n\nPHOTO_FILENAMES = tf.io.gfile.glob(str(GCS_PATH + '/photo_tfrec/*.tfrec'))\nprint('Photo Files:', len(PHOTO_FILENAMES))","metadata":{"execution":{"iopub.status.busy":"2022-04-19T10:49:43.761261Z","iopub.execute_input":"2022-04-19T10:49:43.761663Z","iopub.status.idle":"2022-04-19T10:49:44.326195Z","shell.execute_reply.started":"2022-04-19T10:49:43.761628Z","shell.execute_reply":"2022-04-19T10:49:44.325297Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Zdjęcia z datasetu mają rozmiar 256x256 i są obrazami w formacie RGB - channel = 3. Obrazy są skalowane do przedziału [-1,1].","metadata":{}},{"cell_type":"code","source":"IMAGE_SIZE = [256, 256]\n\ndef decode_image(image):\n    image = tf.image.decode_jpeg(image, channels=3)\n    image = (tf.cast(image, tf.float32) / 127.5) - 1\n    image = tf.reshape(image, [*IMAGE_SIZE, 3])\n    return image\n\ndef read_tfrecord(example):\n    tfrecord_format = {\n        \"image_name\": tf.io.FixedLenFeature([], tf.string),\n        \"image\": tf.io.FixedLenFeature([], tf.string),\n        \"target\": tf.io.FixedLenFeature([], tf.string)\n    }\n    example = tf.io.parse_single_example(example, tfrecord_format)\n    image = decode_image(example['image'])\n    return image","metadata":{"execution":{"iopub.status.busy":"2022-04-19T10:49:51.434041Z","iopub.execute_input":"2022-04-19T10:49:51.434393Z","iopub.status.idle":"2022-04-19T10:49:51.443535Z","shell.execute_reply.started":"2022-04-19T10:49:51.434362Z","shell.execute_reply":"2022-04-19T10:49:51.441339Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def load_dataset(filenames, labeled=True, ordered=False):\n    dataset = tf.data.TFRecordDataset(filenames)\n    dataset = dataset.map(read_tfrecord, num_parallel_calls=AUTOTUNE)\n    return dataset","metadata":{"execution":{"iopub.status.busy":"2022-04-19T10:49:59.976751Z","iopub.execute_input":"2022-04-19T10:49:59.977075Z","iopub.status.idle":"2022-04-19T10:49:59.98187Z","shell.execute_reply.started":"2022-04-19T10:49:59.977043Z","shell.execute_reply":"2022-04-19T10:49:59.980845Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Ładowanie datasetów.","metadata":{}},{"cell_type":"code","source":"monet_ds = load_dataset(MONET_FILENAMES, labeled=True).batch(1)\nphoto_ds = load_dataset(PHOTO_FILENAMES, labeled=True).batch(1)","metadata":{"execution":{"iopub.status.busy":"2022-04-19T10:50:03.891176Z","iopub.execute_input":"2022-04-19T10:50:03.891513Z","iopub.status.idle":"2022-04-19T10:50:06.092927Z","shell.execute_reply.started":"2022-04-19T10:50:03.891481Z","shell.execute_reply":"2022-04-19T10:50:06.092198Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"example_monet = next(iter(monet_ds))\nexample_photo = next(iter(photo_ds))","metadata":{"execution":{"iopub.status.busy":"2022-04-19T10:50:15.5294Z","iopub.execute_input":"2022-04-19T10:50:15.529811Z","iopub.status.idle":"2022-04-19T10:50:16.055146Z","shell.execute_reply.started":"2022-04-19T10:50:15.529765Z","shell.execute_reply":"2022-04-19T10:50:16.054265Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Przykładowe zdjęcie i obraz Moneta.","metadata":{}},{"cell_type":"code","source":"plt.subplot(121)\nplt.title('Photo')\nplt.imshow(example_photo[0] * 0.5 + 0.5)\n\nplt.subplot(122)\nplt.title('Monet')\nplt.imshow(example_monet[0] * 0.5 + 0.5)","metadata":{"execution":{"iopub.status.busy":"2022-04-19T10:50:21.65811Z","iopub.execute_input":"2022-04-19T10:50:21.65847Z","iopub.status.idle":"2022-04-19T10:50:22.379099Z","shell.execute_reply.started":"2022-04-19T10:50:21.658436Z","shell.execute_reply":"2022-04-19T10:50:22.378143Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Generator\n\nŚieć neuronowa jest budowana w architekturze UNET. Definiowane są metody`downsample` and `upsample` methods.\n\nMetoda `downsample` (próbkowanie) pozwala zmniejszyć rozmiar zestawu danych redukując wymiary obrazu. Wysokość i szerokość w wyniku tej operacji pomniejszane są dwukrotnie.","metadata":{}},{"cell_type":"code","source":"OUTPUT_CHANNELS = 3\n\ndef downsample(filters, size, apply_instancenorm=True):\n    initializer = tf.random_normal_initializer(0., 0.02)\n    gamma_init = keras.initializers.RandomNormal(mean=0.0, stddev=0.02)\n\n    result = keras.Sequential()\n    result.add(layers.Conv2D(filters, size, strides=2, padding='same',\n                             kernel_initializer=initializer, use_bias=False))\n\n    if apply_instancenorm:\n        result.add(tfa.layers.InstanceNormalization(gamma_initializer=gamma_init))\n\n    result.add(layers.LeakyReLU())\n\n    return result","metadata":{"execution":{"iopub.status.busy":"2022-04-19T10:50:34.496659Z","iopub.execute_input":"2022-04-19T10:50:34.496967Z","iopub.status.idle":"2022-04-19T10:50:34.50525Z","shell.execute_reply.started":"2022-04-19T10:50:34.496939Z","shell.execute_reply":"2022-04-19T10:50:34.504278Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"`Upsample` to operacja przeciwna do próbkowania, pozwala rozszerzyć dane.","metadata":{}},{"cell_type":"code","source":"def upsample(filters, size, apply_dropout=False):\n    initializer = tf.random_normal_initializer(0., 0.02)\n    gamma_init = keras.initializers.RandomNormal(mean=0.0, stddev=0.02)\n\n    result = keras.Sequential()\n    result.add(layers.Conv2DTranspose(filters, size, strides=2,\n                                      padding='same',\n                                      kernel_initializer=initializer,\n                                      use_bias=False))\n\n    result.add(tfa.layers.InstanceNormalization(gamma_initializer=gamma_init))\n\n    if apply_dropout:\n        result.add(layers.Dropout(0.5))\n\n    result.add(layers.ReLU())\n\n    return result","metadata":{"execution":{"iopub.status.busy":"2022-04-19T10:50:37.843056Z","iopub.execute_input":"2022-04-19T10:50:37.843412Z","iopub.status.idle":"2022-04-19T10:50:37.851309Z","shell.execute_reply.started":"2022-04-19T10:50:37.843379Z","shell.execute_reply":"2022-04-19T10:50:37.850241Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Generator\n\nZadaniem generatora jest na początku wykonie downsamplingu obrazu, a następnie upsamplingu.","metadata":{}},{"cell_type":"code","source":"def Generator():\n    inputs = layers.Input(shape=[256,256,3])\n\n    # bs = batch size\n    down_stack = [\n        downsample(64, 4, apply_instancenorm=False), # (bs, 128, 128, 64)\n        downsample(128, 4), # (bs, 64, 64, 128)\n        downsample(256, 4), # (bs, 32, 32, 256)\n        downsample(512, 4), # (bs, 16, 16, 512)\n        downsample(512, 4), # (bs, 8, 8, 512)\n        downsample(512, 4), # (bs, 4, 4, 512)\n        downsample(512, 4), # (bs, 2, 2, 512)\n        downsample(512, 4), # (bs, 1, 1, 512)\n    ]\n\n    up_stack = [\n        upsample(512, 4, apply_dropout=True), # (bs, 2, 2, 1024)\n        upsample(512, 4, apply_dropout=True), # (bs, 4, 4, 1024)\n        upsample(512, 4, apply_dropout=True), # (bs, 8, 8, 1024)\n        upsample(512, 4), # (bs, 16, 16, 1024)\n        upsample(256, 4), # (bs, 32, 32, 512)\n        upsample(128, 4), # (bs, 64, 64, 256)\n        upsample(64, 4), # (bs, 128, 128, 128)\n    ]\n\n    initializer = tf.random_normal_initializer(0., 0.02)\n    last = layers.Conv2DTranspose(OUTPUT_CHANNELS, 4,\n                                  strides=2,\n                                  padding='same',\n                                  kernel_initializer=initializer,\n                                  activation='tanh') # (bs, 256, 256, 3)\n\n    x = inputs\n\n    # Downsampling through the model\n    skips = []\n    for down in down_stack:\n        x = down(x)\n        skips.append(x)\n\n    skips = reversed(skips[:-1])\n\n    # Upsampling and establishing the skip connections\n    for up, skip in zip(up_stack, skips):\n        x = up(x)\n        x = layers.Concatenate()([x, skip])\n\n    x = last(x)\n\n    return keras.Model(inputs=inputs, outputs=x)","metadata":{"execution":{"iopub.status.busy":"2022-04-19T10:50:45.066655Z","iopub.execute_input":"2022-04-19T10:50:45.066984Z","iopub.status.idle":"2022-04-19T10:50:45.081044Z","shell.execute_reply.started":"2022-04-19T10:50:45.066954Z","shell.execute_reply":"2022-04-19T10:50:45.080124Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Dyskrymiantor\n\nDyskryminator przyjmuje na wejście obraz i klasyfikuje go jako prawdziwy lub fałszywy. Dyskryminator zwraca mniejszy obraz 2D z pikselami o wyższych wartościach wskazujących na \"prawdziwość\", natomiast piksele o niższych wartościach są oznaką \"fałszywości\".","metadata":{}},{"cell_type":"code","source":"def Discriminator():\n    initializer = tf.random_normal_initializer(0., 0.02)\n    gamma_init = keras.initializers.RandomNormal(mean=0.0, stddev=0.02)\n\n    inp = layers.Input(shape=[256, 256, 3], name='input_image')\n\n    x = inp\n\n    down1 = downsample(64, 4, False)(x) # (bs, 128, 128, 64)\n    down2 = downsample(128, 4)(down1) # (bs, 64, 64, 128)\n    down3 = downsample(256, 4)(down2) # (bs, 32, 32, 256)\n\n    zero_pad1 = layers.ZeroPadding2D()(down3) # (bs, 34, 34, 256)\n    conv = layers.Conv2D(512, 4, strides=1,\n                         kernel_initializer=initializer,\n                         use_bias=False)(zero_pad1) # (bs, 31, 31, 512)\n\n    norm1 = tfa.layers.InstanceNormalization(gamma_initializer=gamma_init)(conv)\n\n    leaky_relu = layers.LeakyReLU()(norm1)\n\n    zero_pad2 = layers.ZeroPadding2D()(leaky_relu) # (bs, 33, 33, 512)\n\n    last = layers.Conv2D(1, 4, strides=1,\n                         kernel_initializer=initializer)(zero_pad2) # (bs, 30, 30, 1)\n\n    return tf.keras.Model(inputs=inp, outputs=last)","metadata":{"execution":{"iopub.status.busy":"2022-04-19T10:50:52.657031Z","iopub.execute_input":"2022-04-19T10:50:52.657407Z","iopub.status.idle":"2022-04-19T10:50:52.667866Z","shell.execute_reply.started":"2022-04-19T10:50:52.657374Z","shell.execute_reply":"2022-04-19T10:50:52.666687Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"with strategy.scope():\n    monet_generator = Generator() # transforms photos to Monet-esque paintings\n    photo_generator = Generator() # transforms Monet paintings to be more like photos\n\n    monet_discriminator = Discriminator() # differentiates real Monet paintings and generated Monet paintings\n    photo_discriminator = Discriminator() # differentiates real photos and generated photos","metadata":{"execution":{"iopub.status.busy":"2022-04-19T10:50:57.42602Z","iopub.execute_input":"2022-04-19T10:50:57.426519Z","iopub.status.idle":"2022-04-19T10:50:59.891878Z","shell.execute_reply.started":"2022-04-19T10:50:57.426477Z","shell.execute_reply":"2022-04-19T10:50:59.89113Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Generator na tym etapie nie jest wytrenowany, w związku z czym wygenerowane obrazy nie przypominają dzieł Moneta.","metadata":{}},{"cell_type":"code","source":"to_monet = monet_generator(example_photo)\n\nplt.subplot(1, 2, 1)\nplt.title(\"Original Photo\")\nplt.imshow(example_photo[0] * 0.5 + 0.5)\n\nplt.subplot(1, 2, 2)\nplt.title(\"Monet-esque Photo\")\nplt.imshow(to_monet[0] * 0.5 + 0.5)\nplt.show()","metadata":{"execution":{"iopub.status.busy":"2022-04-19T10:51:04.737736Z","iopub.execute_input":"2022-04-19T10:51:04.738049Z","iopub.status.idle":"2022-04-19T10:51:07.653987Z","shell.execute_reply.started":"2022-04-19T10:51:04.738019Z","shell.execute_reply":"2022-04-19T10:51:07.653076Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# CycleGAN\n\nPodczas trenowania model będzie zamianiał zdjęcie na obraz Moneta i później stworzony obraz z powrotem na zdjęcie. Różnica między oryginalnem obrazem i tym, który powstał po podwójnym przekształceniu stanie się funkcją kosztu spójności między cyklami. W idealnym scenariuszu dwukrotnie przetransformowane zdjęcie i oryginalne będą identyczne.","metadata":{}},{"cell_type":"code","source":"class CycleGan(keras.Model):\n    def __init__(\n        self,\n        monet_generator,\n        photo_generator,\n        monet_discriminator,\n        photo_discriminator,\n        lambda_cycle=10,\n    ):\n        super(CycleGan, self).__init__()\n        self.m_gen = monet_generator\n        self.p_gen = photo_generator\n        self.m_disc = monet_discriminator\n        self.p_disc = photo_discriminator\n        self.lambda_cycle = lambda_cycle\n        \n    def compile(\n        self,\n        m_gen_optimizer,\n        p_gen_optimizer,\n        m_disc_optimizer,\n        p_disc_optimizer,\n        gen_loss_fn,\n        disc_loss_fn,\n        cycle_loss_fn,\n        identity_loss_fn\n    ):\n        super(CycleGan, self).compile()\n        self.m_gen_optimizer = m_gen_optimizer\n        self.p_gen_optimizer = p_gen_optimizer\n        self.m_disc_optimizer = m_disc_optimizer\n        self.p_disc_optimizer = p_disc_optimizer\n        self.gen_loss_fn = gen_loss_fn\n        self.disc_loss_fn = disc_loss_fn\n        self.cycle_loss_fn = cycle_loss_fn\n        self.identity_loss_fn = identity_loss_fn\n        \n    def train_step(self, batch_data):\n        real_monet, real_photo = batch_data\n        \n        with tf.GradientTape(persistent=True) as tape:\n            # photo to monet back to photo\n            fake_monet = self.m_gen(real_photo, training=True)\n            cycled_photo = self.p_gen(fake_monet, training=True)\n\n            # monet to photo back to monet\n            fake_photo = self.p_gen(real_monet, training=True)\n            cycled_monet = self.m_gen(fake_photo, training=True)\n\n            # generating itself\n            same_monet = self.m_gen(real_monet, training=True)\n            same_photo = self.p_gen(real_photo, training=True)\n\n            # discriminator used to check, inputing real images\n            disc_real_monet = self.m_disc(real_monet, training=True)\n            disc_real_photo = self.p_disc(real_photo, training=True)\n\n            # discriminator used to check, inputing fake images\n            disc_fake_monet = self.m_disc(fake_monet, training=True)\n            disc_fake_photo = self.p_disc(fake_photo, training=True)\n\n            # evaluates generator loss\n            monet_gen_loss = self.gen_loss_fn(disc_fake_monet)\n            photo_gen_loss = self.gen_loss_fn(disc_fake_photo)\n\n            # evaluates total cycle consistency loss\n            total_cycle_loss = self.cycle_loss_fn(real_monet, cycled_monet, self.lambda_cycle) + self.cycle_loss_fn(real_photo, cycled_photo, self.lambda_cycle)\n\n            # evaluates total generator loss\n            total_monet_gen_loss = monet_gen_loss + total_cycle_loss + self.identity_loss_fn(real_monet, same_monet, self.lambda_cycle)\n            total_photo_gen_loss = photo_gen_loss + total_cycle_loss + self.identity_loss_fn(real_photo, same_photo, self.lambda_cycle)\n\n            # evaluates discriminator loss\n            monet_disc_loss = self.disc_loss_fn(disc_real_monet, disc_fake_monet)\n            photo_disc_loss = self.disc_loss_fn(disc_real_photo, disc_fake_photo)\n\n        # Calculate the gradients for generator and discriminator\n        monet_generator_gradients = tape.gradient(total_monet_gen_loss,\n                                                  self.m_gen.trainable_variables)\n        photo_generator_gradients = tape.gradient(total_photo_gen_loss,\n                                                  self.p_gen.trainable_variables)\n\n        monet_discriminator_gradients = tape.gradient(monet_disc_loss,\n                                                      self.m_disc.trainable_variables)\n        photo_discriminator_gradients = tape.gradient(photo_disc_loss,\n                                                      self.p_disc.trainable_variables)\n\n        # Apply the gradients to the optimizer\n        self.m_gen_optimizer.apply_gradients(zip(monet_generator_gradients,\n                                                 self.m_gen.trainable_variables))\n\n        self.p_gen_optimizer.apply_gradients(zip(photo_generator_gradients,\n                                                 self.p_gen.trainable_variables))\n\n        self.m_disc_optimizer.apply_gradients(zip(monet_discriminator_gradients,\n                                                  self.m_disc.trainable_variables))\n\n        self.p_disc_optimizer.apply_gradients(zip(photo_discriminator_gradients,\n                                                  self.p_disc.trainable_variables))\n        \n        return {\n            \"monet_gen_loss\": total_monet_gen_loss,\n            \"photo_gen_loss\": total_photo_gen_loss,\n            \"monet_disc_loss\": monet_disc_loss,\n            \"photo_disc_loss\": photo_disc_loss\n        }","metadata":{"execution":{"iopub.status.busy":"2022-04-19T10:51:19.571587Z","iopub.execute_input":"2022-04-19T10:51:19.571948Z","iopub.status.idle":"2022-04-19T10:51:19.596647Z","shell.execute_reply.started":"2022-04-19T10:51:19.571916Z","shell.execute_reply":"2022-04-19T10:51:19.595442Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Loss functions - funkcja kosztu\n\nFunkcja kosztu w dyskryminatorze porównuje oryginalne obrazy do macierzy wypełnionej 1, a fałszywe - 0. Idealny dyskrymiantor da na wyjściu same 1 dla obrazu prawdziwego i macierz zer dla fałszywego.","metadata":{}},{"cell_type":"code","source":"with strategy.scope():\n    def discriminator_loss(real, generated):\n        real_loss = tf.keras.losses.BinaryCrossentropy(from_logits=True, reduction=tf.keras.losses.Reduction.NONE)(tf.ones_like(real), real)\n\n        generated_loss = tf.keras.losses.BinaryCrossentropy(from_logits=True, reduction=tf.keras.losses.Reduction.NONE)(tf.zeros_like(generated), generated)\n\n        total_disc_loss = real_loss + generated_loss\n\n        return total_disc_loss * 0.5","metadata":{"execution":{"iopub.status.busy":"2022-04-19T10:51:27.016276Z","iopub.execute_input":"2022-04-19T10:51:27.016626Z","iopub.status.idle":"2022-04-19T10:51:27.026075Z","shell.execute_reply.started":"2022-04-19T10:51:27.016597Z","shell.execute_reply":"2022-04-19T10:51:27.025183Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Generator próbuje oszukać dyskryminator generując obraz tak, by dyskryminator uznał go za oryginalny. Idealny generator spowoduje, że dyskryminator na wyjściu poda macierz wypełnioną 1.","metadata":{}},{"cell_type":"code","source":"with strategy.scope():\n    def generator_loss(generated):\n        return tf.keras.losses.BinaryCrossentropy(from_logits=True, reduction=tf.keras.losses.Reduction.NONE)(tf.ones_like(generated), generated)","metadata":{"execution":{"iopub.status.busy":"2022-04-19T10:51:31.305893Z","iopub.execute_input":"2022-04-19T10:51:31.306278Z","iopub.status.idle":"2022-04-19T10:51:31.311704Z","shell.execute_reply.started":"2022-04-19T10:51:31.306245Z","shell.execute_reply":"2022-04-19T10:51:31.310838Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"import tensorflow as tf\nfrom tensorflow import keras\nfrom tensorflow.keras import layers\nimport tensorflow_addons as tfa\n\nfrom kaggle_datasets import KaggleDatasets\nimport matplotlib.pyplot as plt\nimport numpy as np\n\ntry:\n    tpu = tf.distribute.cluster_resolver.TPUClusterResolver()\n    print('Device:', tpu.master())\n    tf.config.experimental_connect_to_cluster(tpu)\n    tf.tpu.experimental.initialize_tpu_system(tpu)\n    strategy = tf.distribute.experimental.TPUStrategy(tpu)\nexcept:\n    strategy = tf.distribute.get_strategy()\nprint('Number of replicas:', strategy.num_replicas_in_sync)\n\nAUTOTUNE = tf.data.experimental.AUTOTUNE\n    \nprint(tf.__version__)","metadata":{"_uuid":"d629ff2d2480ee46fbb7e2d37f6b5fab8052498a","_cell_guid":"79c7e3d0-c299-4dcb-8224-4455121ee9b0","execution":{"iopub.status.busy":"2022-04-19T10:49:07.292917Z","iopub.execute_input":"2022-04-19T10:49:07.293257Z","iopub.status.idle":"2022-04-19T10:49:11.583009Z","shell.execute_reply.started":"2022-04-19T10:49:07.293223Z","shell.execute_reply":"2022-04-19T10:49:11.582139Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Cycle consistency loss to średnia arytmetyczna z różnic między pierwotnym zdjęciem i dwukrotnie przetransformowanym.","metadata":{}},{"cell_type":"code","source":"with strategy.scope():\n    def calc_cycle_loss(real_image, cycled_image, LAMBDA):\n        loss1 = tf.reduce_mean(tf.abs(real_image - cycled_image))\n\n        return LAMBDA * loss1","metadata":{"execution":{"iopub.status.busy":"2022-04-19T10:51:37.467694Z","iopub.execute_input":"2022-04-19T10:51:37.468039Z","iopub.status.idle":"2022-04-19T10:51:37.473849Z","shell.execute_reply.started":"2022-04-19T10:51:37.468008Z","shell.execute_reply":"2022-04-19T10:51:37.472904Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Identity loss służy do porównywania obrazu x i generatora tego obrazu F, który produkuje obrazy.\nOczekujemy, że F(x)~x, czyli jeśli go generatora obrazów w stylu Moneta poda się obraz Moneta,\nna wyjściu powinno się otrzymać ten sam obraz.","metadata":{}},{"cell_type":"code","source":"with strategy.scope():\n    def identity_loss(real_image, same_image, LAMBDA):\n        loss = tf.reduce_mean(tf.abs(real_image - same_image))\n        return LAMBDA * 0.5 * loss","metadata":{"execution":{"iopub.status.busy":"2022-04-19T10:51:41.584636Z","iopub.execute_input":"2022-04-19T10:51:41.58497Z","iopub.status.idle":"2022-04-19T10:51:41.589805Z","shell.execute_reply.started":"2022-04-19T10:51:41.584938Z","shell.execute_reply":"2022-04-19T10:51:41.588807Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Trenowanie CycleGAN\n\nModel jest kompilowany, a następnie trenowany za pomocą wywołania funkcji `fit`.","metadata":{}},{"cell_type":"code","source":"with strategy.scope():\n    monet_generator_optimizer = tf.keras.optimizers.Adam(2e-4, beta_1=0.5)\n    photo_generator_optimizer = tf.keras.optimizers.Adam(2e-4, beta_1=0.5)\n\n    monet_discriminator_optimizer = tf.keras.optimizers.Adam(2e-4, beta_1=0.5)\n    photo_discriminator_optimizer = tf.keras.optimizers.Adam(2e-4, beta_1=0.5)","metadata":{"execution":{"iopub.status.busy":"2022-04-19T10:51:45.12875Z","iopub.execute_input":"2022-04-19T10:51:45.129188Z","iopub.status.idle":"2022-04-19T10:51:45.135853Z","shell.execute_reply.started":"2022-04-19T10:51:45.129123Z","shell.execute_reply":"2022-04-19T10:51:45.134693Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"with strategy.scope():\n    cycle_gan_model = CycleGan(\n        monet_generator, photo_generator, monet_discriminator, photo_discriminator\n    )\n\n    cycle_gan_model.compile(\n        m_gen_optimizer = monet_generator_optimizer,\n        p_gen_optimizer = photo_generator_optimizer,\n        m_disc_optimizer = monet_discriminator_optimizer,\n        p_disc_optimizer = photo_discriminator_optimizer,\n        gen_loss_fn = generator_loss,\n        disc_loss_fn = discriminator_loss,\n        cycle_loss_fn = calc_cycle_loss,\n        identity_loss_fn = identity_loss\n    )","metadata":{"execution":{"iopub.status.busy":"2022-04-19T10:51:48.790243Z","iopub.execute_input":"2022-04-19T10:51:48.790578Z","iopub.status.idle":"2022-04-19T10:51:48.816719Z","shell.execute_reply.started":"2022-04-19T10:51:48.790548Z","shell.execute_reply":"2022-04-19T10:51:48.816015Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"cycle_gan_model.fit(\n    tf.data.Dataset.zip((monet_ds, photo_ds)),\n    epochs=25\n)","metadata":{"execution":{"iopub.status.busy":"2022-04-19T10:51:53.755943Z","iopub.execute_input":"2022-04-19T10:51:53.756273Z","iopub.status.idle":"2022-04-19T10:58:03.532082Z","shell.execute_reply.started":"2022-04-19T10:51:53.756243Z","shell.execute_reply":"2022-04-19T10:58:03.5302Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Efekty","metadata":{}},{"cell_type":"code","source":"_, ax = plt.subplots(5, 2, figsize=(12, 12))\nfor i, img in enumerate(photo_ds.take(5)):\n    prediction = monet_generator(img, training=False)[0].numpy()\n    prediction = (prediction * 127.5 + 127.5).astype(np.uint8)\n    img = (img[0] * 127.5 + 127.5).numpy().astype(np.uint8)\n\n    ax[i, 0].imshow(img)\n    ax[i, 1].imshow(prediction)\n    ax[i, 0].set_title(\"Input Photo\")\n    ax[i, 1].set_title(\"Monet-esque\")\n    ax[i, 0].axis(\"off\")\n    ax[i, 1].axis(\"off\")\nplt.show()","metadata":{"execution":{"iopub.status.busy":"2022-04-19T10:58:17.685488Z","iopub.execute_input":"2022-04-19T10:58:17.685814Z","iopub.status.idle":"2022-04-19T10:58:18.852615Z","shell.execute_reply.started":"2022-04-19T10:58:17.685783Z","shell.execute_reply":"2022-04-19T10:58:18.850697Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Zapisywanie obrazów","metadata":{}},{"cell_type":"code","source":"import PIL\n! mkdir ../images","metadata":{"execution":{"iopub.status.busy":"2022-04-19T13:55:31.157402Z","iopub.execute_input":"2022-04-19T13:55:31.158436Z","iopub.status.idle":"2022-04-19T13:55:31.964693Z","shell.execute_reply.started":"2022-04-19T13:55:31.158386Z","shell.execute_reply":"2022-04-19T13:55:31.963453Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"i = 1\nfor img in photo_ds:\n    prediction = monet_generator(img, training=False)[0].numpy()\n    prediction = (prediction * 127.5 + 127.5).astype(np.uint8)\n    im = PIL.Image.fromarray(prediction)\n    im.save(\"../images/\" + str(i) + \".jpg\")\n    i += 1","metadata":{"execution":{"iopub.status.busy":"2022-04-19T13:55:34.35824Z","iopub.execute_input":"2022-04-19T13:55:34.358604Z","iopub.status.idle":"2022-04-19T13:55:34.43197Z","shell.execute_reply.started":"2022-04-19T13:55:34.35857Z","shell.execute_reply":"2022-04-19T13:55:34.430569Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"import shutil\nshutil.make_archive(\"/kaggle/working/images\", 'zip', \"/kaggle/images\")","metadata":{"trusted":true},"execution_count":null,"outputs":[]}]}