{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"name":"python","version":"3.7.12","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"},"kaggle":{"accelerator":"gpu","dataSources":[{"sourceId":21755,"databundleVersionId":1475600,"sourceType":"competition"}],"dockerImageVersionId":30302,"isInternetEnabled":true,"language":"python","sourceType":"notebook","isGpuEnabled":true}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"markdown","source":"## Dependencies","metadata":{}},{"cell_type":"code","source":"import os, random, json, PIL, shutil, re\nimport numpy as np\nimport pandas as pd\nimport seaborn as sns\nimport matplotlib.pyplot as plt\nfrom kaggle_datasets import KaggleDatasets\nimport tensorflow as tf\nimport tensorflow.keras.layers as L\nimport tensorflow_addons as tfa\nfrom tensorflow.keras import Model, losses, optimizers","metadata":{"_uuid":"d629ff2d2480ee46fbb7e2d37f6b5fab8052498a","_cell_guid":"79c7e3d0-c299-4dcb-8224-4455121ee9b0","trusted":true,"_kg_hide-input":true,"execution":{"iopub.status.busy":"2024-12-09T01:20:54.673145Z","iopub.execute_input":"2024-12-09T01:20:54.673548Z","iopub.status.idle":"2024-12-09T01:21:00.697237Z","shell.execute_reply.started":"2024-12-09T01:20:54.673472Z","shell.execute_reply":"2024-12-09T01:21:00.696288Z"}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"## TPU configuration","metadata":{}},{"cell_type":"code","source":"try:\n    tpu = tf.distribute.cluster_resolver.TPUClusterResolver()\n    print(f'Running on TPU {tpu.master()}')\nexcept ValueError:\n    tpu = None\n\nif tpu:\n    tf.config.experimental_connect_to_cluster(tpu)\n    tf.tpu.experimental.initialize_tpu_system(tpu)\n    strategy = tf.distribute.experimental.TPUStrategy(tpu)\nelse:\n    strategy = tf.distribute.get_strategy()\n\n\nREPLICAS = strategy.num_replicas_in_sync\nprint(f'REPLICAS: {REPLICAS}')\nAUTO = tf.data.experimental.AUTOTUNE","metadata":{"_kg_hide-input":true,"trusted":true,"execution":{"iopub.status.busy":"2024-12-09T01:21:00.698867Z","iopub.execute_input":"2024-12-09T01:21:00.699355Z","iopub.status.idle":"2024-12-09T01:21:00.710579Z","shell.execute_reply.started":"2024-12-09T01:21:00.699328Z","shell.execute_reply":"2024-12-09T01:21:00.709712Z"}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"# Model parameters","metadata":{}},{"cell_type":"code","source":"HEIGHT = 256\nWIDTH = 256\nCHANNELS = 3\nEPOCHS = 5\nBATCH_SIZE = 2\nAUTOTUNE = tf.data.experimental.AUTOTUNE","metadata":{"_kg_hide-input":false,"trusted":true,"execution":{"iopub.status.busy":"2024-12-09T01:21:00.711541Z","iopub.execute_input":"2024-12-09T01:21:00.711894Z","iopub.status.idle":"2024-12-09T01:21:00.720442Z","shell.execute_reply.started":"2024-12-09T01:21:00.711856Z","shell.execute_reply":"2024-12-09T01:21:00.719571Z"}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"# Load data","metadata":{}},{"cell_type":"code","source":"GCS_PATH = KaggleDatasets().get_gcs_path('monet-gan-getting-started')\n\nMONET_FILENAMES = tf.io.gfile.glob(str(GCS_PATH + '/monet_tfrec/*.tfrec'))\nPHOTO_FILENAMES = tf.io.gfile.glob(str(GCS_PATH + '/photo_tfrec/*.tfrec'))\n\ndef count_data_items(filenames):\n    n = [int(re.compile(r\"-([0-9]*)\\.\").search(filename).group(1)) for filename in filenames]\n    return np.sum(n)\n\nn_monet_samples = count_data_items(MONET_FILENAMES)*4\nn_photo_samples = count_data_items(PHOTO_FILENAMES)\n\nprint(f'Monet TFRecord files: {len(MONET_FILENAMES)}')\nprint(f'Monet image files: {n_monet_samples}')\nprint(f'Photo TFRecord files: {len(PHOTO_FILENAMES)}')\nprint(f'Photo image files: {n_photo_samples}')","metadata":{"trusted":true,"_kg_hide-input":true,"execution":{"iopub.status.busy":"2024-12-09T01:21:00.722519Z","iopub.execute_input":"2024-12-09T01:21:00.722764Z","iopub.status.idle":"2024-12-09T01:21:04.001246Z","shell.execute_reply.started":"2024-12-09T01:21:00.722742Z","shell.execute_reply":"2024-12-09T01:21:04.000235Z"}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"## Auxiliar functions","metadata":{}},{"cell_type":"code","source":"def decode_image(image):\n    image = tf.image.decode_jpeg(image, channels=CHANNELS)\n    image = (tf.cast(image, tf.float32) / 127.5) - 1\n    image = tf.reshape(image, [HEIGHT, WIDTH, CHANNELS])\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\n\ndef flip(image):\n    return tf.image.flip_left_right(image)\n\ndef random_crop(image):\n    cropped_image = tf.image.random_crop(image, size=[256, 256, 3])\n    return cropped_image\n\ndef random_jitter(image):\n    # resizing to 286 x 286 x 3 \n    image = tf.image.resize(image, [int(256*1.3), int(256*1.3)],\n                          method=tf.image.ResizeMethod.NEAREST_NEIGHBOR)\n    # randomly cropping to 256 x 256 x 3\n    image = random_crop(image)\n    # random mirroring\n    return image\n\ndef load_dataset(filenames,Is_Monet=False):\n    dataset = tf.data.TFRecordDataset(filenames)\n    dataset = dataset.map(read_tfrecord, num_parallel_calls=AUTO)\n    if Is_Monet:\n        dataset = dataset.concatenate(dataset.map(flip, num_parallel_calls=AUTOTUNE).shuffle(100000))\n        dataset = dataset.concatenate(dataset.map(random_jitter, num_parallel_calls=AUTOTUNE).shuffle(10000, reshuffle_each_iteration=True))\n    num_elements = 0\n    for element in dataset:\n        num_elements += 1\n    print(num_elements)\n    return dataset\n\ndef get_gan_dataset(monet_files, photo_files, augment=None, repeat=True, shuffle=True, batch_size=1):\n\n    monet_ds = load_dataset(monet_files,Is_Monet=True)\n    photo_ds = load_dataset(photo_files,Is_Monet=False)\n\n    if repeat:\n        monet_ds = monet_ds.repeat()\n        photo_ds = photo_ds.repeat()\n    if shuffle:\n        monet_ds = monet_ds.shuffle(2048)\n        photo_ds = photo_ds.shuffle(2048)\n        \n    monet_ds = monet_ds.batch(batch_size, drop_remainder=True)\n    photo_ds = photo_ds.batch(batch_size, drop_remainder=True)\n    monet_ds = monet_ds.cache()\n    photo_ds = photo_ds.cache()\n    monet_ds = monet_ds.prefetch(AUTO)\n    photo_ds = photo_ds.prefetch(AUTO)\n    \n    gan_ds = tf.data.Dataset.zip((monet_ds, photo_ds))\n    \n    return gan_ds\n\ndef display_samples(ds, row, col):\n    ds_iter = iter(ds)\n    plt.figure(figsize=(15, int(15*row/col)))\n    for j in range(row*col):\n        example_sample = next(ds_iter)\n        plt.subplot(row,col,j+1)\n        plt.axis('off')\n        plt.imshow(example_sample[0] * 0.5 + 0.5)\n    plt.show()\n        \ndef display_generated_samples(ds, model, n_samples):\n    ds_iter = iter(ds)\n    for n_sample in range(n_samples):\n        example_sample = next(ds_iter)\n        generated_sample = model.predict(example_sample)\n        \n        plt.subplot(121)\n        plt.title(\"input image\")\n        plt.imshow(example_sample[0] * 0.5 + 0.5)\n        plt.axis('off')\n        plt.subplot(122)\n        plt.title(\"Generated image\")\n        plt.imshow(generated_sample[0] * 0.5 + 0.5)\n        plt.axis('off')\n        plt.show()\n        \ndef predict_and_save(input_ds, generator_model, output_path):\n    i = 1\n    for img in input_ds:\n        prediction = generator_model(img, training=False)[0].numpy() # make predition\n        prediction = (prediction * 127.5 + 127.5).astype(np.uint8)   # re-scale\n        im = PIL.Image.fromarray(prediction)\n        im.save(f'{output_path}{str(i)}.jpg')\n        i += 1\n                \n\n# Model functions\ndef downsample(filters, size, apply_instancenorm=True, strides=2):\n    initializer = tf.random_normal_initializer(0., 0.02)\n    gamma_init = tf.keras.initializers.RandomNormal(mean=0.0, stddev=0.02)\n\n    result = tf.keras.Sequential()\n    result.add(L.Conv2D(filters, size, strides=strides, 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(L.LeakyReLU())\n\n    return result\n\ndef upsample(filters, size, apply_dropout=False, strides=2):\n    initializer = tf.random_normal_initializer(0., 0.02)\n    gamma_init = tf.keras.initializers.RandomNormal(mean=0.0, stddev=0.02)\n\n    result = tf.keras.Sequential()\n    result.add(L.Conv2DTranspose(filters, size, strides=strides, padding='same',\n                                 kernel_initializer=initializer, use_bias=False))\n\n    result.add(tfa.layers.InstanceNormalization(gamma_initializer=gamma_init))\n\n    if apply_dropout:\n        result.add(L.Dropout(0.5))\n\n    result.add(L.ReLU())\n\n    return result","metadata":{"trusted":true,"_kg_hide-input":true,"execution":{"iopub.status.busy":"2024-12-09T01:21:04.002834Z","iopub.execute_input":"2024-12-09T01:21:04.003199Z","iopub.status.idle":"2024-12-09T01:21:04.023520Z","shell.execute_reply.started":"2024-12-09T01:21:04.003163Z","shell.execute_reply":"2024-12-09T01:21:04.022626Z"}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"## Looking at a few Monet paintings","metadata":{}},{"cell_type":"code","source":"display_samples(load_dataset(MONET_FILENAMES,Is_Monet=True).batch(1), 4, 6)","metadata":{"_kg_hide-input":true,"trusted":true,"execution":{"iopub.status.busy":"2024-12-09T01:21:04.024850Z","iopub.execute_input":"2024-12-09T01:21:04.025179Z","iopub.status.idle":"2024-12-09T01:22:02.715415Z","shell.execute_reply.started":"2024-12-09T01:21:04.025142Z","shell.execute_reply":"2024-12-09T01:22:02.714556Z"}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"## Looking at a few photo samples","metadata":{}},{"cell_type":"code","source":"display_samples(load_dataset(PHOTO_FILENAMES,Is_Monet=False).batch(1), 4, 6)","metadata":{"_kg_hide-input":true,"trusted":true,"execution":{"iopub.status.busy":"2024-12-09T01:22:02.716763Z","iopub.execute_input":"2024-12-09T01:22:02.717088Z","iopub.status.idle":"2024-12-09T01:23:12.924973Z","shell.execute_reply.started":"2024-12-09T01:22:02.717055Z","shell.execute_reply":"2024-12-09T01:23:12.924001Z"}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"# Generator model","metadata":{}},{"cell_type":"code","source":"OUTPUT_CHANNELS = 3\n\ndef generator_fn():\n    inputs = L.Input(shape=[HEIGHT, WIDTH, CHANNELS])\n\n    initializer = tf.random_normal_initializer(0., 0.02)\n\n    # Define downsampling layers with self-attention mechanism\n    def downsample(filters, size, apply_instancenorm=True):\n        result = tf.keras.Sequential()\n        result.add(L.Conv2D(filters, size, strides=2, padding='same', \n                            kernel_initializer=initializer, use_bias=False))\n        if apply_instancenorm:\n            result.add(tfa.layers.InstanceNormalization())\n        result.add(L.LeakyReLU())\n        return result\n\n    # Define upsampling layers with improved upsampling strategy\n    def upsample(filters, size, apply_dropout=False):\n        result = tf.keras.Sequential()\n        result.add(L.Conv2DTranspose(filters, size, strides=2, padding='same', \n                                     kernel_initializer=initializer, use_bias=False))\n        result.add(tfa.layers.InstanceNormalization())\n        if apply_dropout:\n            result.add(L.Dropout(0.5))\n        result.add(L.ReLU())\n        return result\n\n    # Define a self-attention block\n    def self_attention(x):\n        channels = x.shape[-1]\n    \n        # Query, Key, and Value transformations\n        f = L.Conv2D(channels // 8, kernel_size=1)(x)  # Query\n        g = L.Conv2D(channels // 8, kernel_size=1)(x)  # Key\n        h = L.Conv2D(channels, kernel_size=1)(x)       # Value\n    \n        # Reshape tensors for proper alignment in the Dot layers\n        f = L.Reshape((-1, channels // 8))(f)  # (batch_size, height * width, channels // 8)\n        g = L.Reshape((-1, channels // 8))(g)  # (batch_size, height * width, channels // 8)\n        h = L.Reshape((-1, channels))(h)       # (batch_size, height * width, channels)\n    \n        # Compute attention map\n        attention_map = L.Softmax(axis=-1)(L.Dot(axes=[2, 2])([f, g]))  # (batch_size, height * width, height * width)\n    \n        # Apply attention map to Value tensor\n        attention = L.Dot(axes=[2, 1])([attention_map, h])  # (batch_size, height * width, channels)\n    \n        # Reshape back to original dimensions\n        attention = L.Reshape(x.shape[1:])(attention)  # (batch_size, height, width, channels)\n    \n        return L.Add()([x, attention])  # Residual connection\n\n\n    # Downsample stack\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    # Upsample stack\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    # Define the last layer\n    last = L.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        # Add self-attention to intermediate layers\n        if x.shape[1] in [16, 8]:  # Apply attention at specific spatial resolutions\n            x = self_attention(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 = L.Concatenate()([x, skip])\n\n    x = last(x)\n\n    return Model(inputs=inputs, outputs=x)","metadata":{"trusted":true,"_kg_hide-input":false,"execution":{"iopub.status.busy":"2024-12-09T01:23:12.926412Z","iopub.execute_input":"2024-12-09T01:23:12.926664Z","iopub.status.idle":"2024-12-09T01:23:12.942216Z","shell.execute_reply.started":"2024-12-09T01:23:12.926636Z","shell.execute_reply":"2024-12-09T01:23:12.941316Z"}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"# Discriminator model","metadata":{}},{"cell_type":"code","source":"def discriminator_fn():\n    initializer = tf.random_normal_initializer(0., 0.02)\n\n    def conv_block(x, filters, kernel_size, strides=1, apply_norm=True, apply_dropout=False):\n        x = L.Conv2D(filters, kernel_size, strides=strides, padding='same',\n                     kernel_initializer=initializer, use_bias=not apply_norm)(x)\n        if apply_norm:\n            x = tfa.layers.InstanceNormalization()(x)\n        x = L.LeakyReLU(alpha=0.2)(x)\n        if apply_dropout:\n            x = L.Dropout(0.3)(x)\n        return x\n\n    inp = L.Input(shape=[HEIGHT, WIDTH, CHANNELS], name='input_image')\n\n    # Downsample blocks\n    x = conv_block(inp, 64, kernel_size=4, strides=2, apply_norm=False)  # (bs, 128, 128, 64)\n    x = conv_block(x, 128, kernel_size=4, strides=2)                     # (bs, 64, 64, 128)\n    x = conv_block(x, 256, kernel_size=4, strides=2)                     # (bs, 32, 32, 256)\n    x = conv_block(x, 512, kernel_size=4, strides=1)                     # (bs, 31, 31, 512)\n\n    # Final convolution to output a single channel\n    x = L.ZeroPadding2D()(x)                                             # (bs, 33, 33, 512)\n    last = L.Conv2D(1, kernel_size=4, strides=1,\n                    kernel_initializer=initializer)(x)                   # (bs, 30, 30, 1)\n\n    return Model(inputs=inp, outputs=last, name='Discriminator')\n","metadata":{"trusted":true,"_kg_hide-input":false,"execution":{"iopub.status.busy":"2024-12-09T01:23:12.943243Z","iopub.execute_input":"2024-12-09T01:23:12.943566Z","iopub.status.idle":"2024-12-09T01:23:12.959816Z","shell.execute_reply.started":"2024-12-09T01:23:12.943535Z","shell.execute_reply":"2024-12-09T01:23:12.959132Z"}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"# Build model (CycleGAN)","metadata":{}},{"cell_type":"code","source":"with strategy.scope():\n    monet_generator = generator_fn() # transforms photos to Monet-esque paintings\n    photo_generator = generator_fn() # transforms Monet paintings to be more like photos\n\n    monet_discriminator = discriminator_fn() # differentiates real Monet paintings and generated Monet paintings\n    photo_discriminator = discriminator_fn() # differentiates real photos and generated photos\n\n\nclass CycleGan(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            # Generator outputs\n            fake_monet = self.m_gen(real_photo, training=True)\n            cycled_photo = self.p_gen(fake_monet, training=True)\n            fake_photo = self.p_gen(real_monet, training=True)\n            cycled_monet = self.m_gen(fake_photo, training=True)\n    \n            # Identity mapping\n            same_monet = self.m_gen(real_monet, training=True)\n            same_photo = self.p_gen(real_photo, training=True)\n    \n            # Discriminator outputs\n            disc_real_monet = self.m_disc(real_monet, training=True)\n            disc_real_photo = self.p_disc(real_photo, training=True)\n            disc_fake_monet = self.m_disc(fake_monet, training=True)\n            disc_fake_photo = self.p_disc(fake_photo, training=True)\n    \n            # Generator losses\n            monet_gen_loss = self.gen_loss_fn(disc_fake_monet)\n            photo_gen_loss = self.gen_loss_fn(disc_fake_photo)\n            total_cycle_loss = self.cycle_loss_fn(real_monet, cycled_monet, self.lambda_cycle) + \\\n                               self.cycle_loss_fn(real_photo, cycled_photo, self.lambda_cycle)\n            total_monet_gen_loss = monet_gen_loss + total_cycle_loss + \\\n                                   self.identity_loss_fn(real_monet, same_monet, self.lambda_cycle)\n            total_photo_gen_loss = photo_gen_loss + total_cycle_loss + \\\n                                   self.identity_loss_fn(real_photo, same_photo, self.lambda_cycle)\n    \n            # Perceptual loss\n            if hasattr(self, \"vgg_model\"):\n                perceptual_loss_monet = perceptual_loss(self.vgg_model, real_monet, fake_monet)\n                perceptual_loss_photo = perceptual_loss(self.vgg_model, real_photo, fake_photo)\n                total_monet_gen_loss += perceptual_loss_monet\n                total_photo_gen_loss += perceptual_loss_photo\n    \n            # Discriminator losses\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        # Apply gradients\n        monet_generator_gradients = tape.gradient(total_monet_gen_loss, self.m_gen.trainable_variables)\n        photo_generator_gradients = tape.gradient(total_photo_gen_loss, self.p_gen.trainable_variables)\n        monet_discriminator_gradients = tape.gradient(monet_disc_loss, self.m_disc.trainable_variables)\n        photo_discriminator_gradients = tape.gradient(photo_disc_loss, self.p_disc.trainable_variables)\n    \n        self.m_gen_optimizer.apply_gradients(zip(monet_generator_gradients, self.m_gen.trainable_variables))\n        self.p_gen_optimizer.apply_gradients(zip(photo_generator_gradients, self.p_gen.trainable_variables))\n        self.m_disc_optimizer.apply_gradients(zip(monet_discriminator_gradients, self.m_disc.trainable_variables))\n        self.p_disc_optimizer.apply_gradients(zip(photo_discriminator_gradients, 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        }\n","metadata":{"trusted":true,"_kg_hide-input":true,"execution":{"iopub.status.busy":"2024-12-09T01:23:35.525239Z","iopub.execute_input":"2024-12-09T01:23:35.525613Z","iopub.status.idle":"2024-12-09T01:23:37.802036Z","shell.execute_reply.started":"2024-12-09T01:23:35.525581Z","shell.execute_reply":"2024-12-09T01:23:37.801081Z"}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"# Loss functions","metadata":{}},{"cell_type":"code","source":"with strategy.scope():\n    # Discriminator loss {0: fake, 1: real} (The discriminator loss outputs the average of the real and generated loss)\n    def discriminator_loss(real, generated):\n        real_loss = losses.BinaryCrossentropy(from_logits=True, reduction=losses.Reduction.NONE)(tf.ones_like(real), real)\n\n        generated_loss = losses.BinaryCrossentropy(from_logits=True, reduction=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\n    \n    # Generator loss\n    def generator_loss(generated):\n        \n        return losses.BinaryCrossentropy(from_logits=True, reduction=losses.Reduction.NONE)(tf.ones_like(generated), generated)\n    \n    \n    # Cycle consistency loss (measures if original photo and the twice transformed photo to be similar to one another)\n    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\n\n    # Identity loss (compares the image with its generator (i.e. photo with photo generator))\n    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":{"trusted":true,"_kg_hide-input":true,"execution":{"iopub.status.busy":"2024-12-09T01:23:55.882839Z","iopub.execute_input":"2024-12-09T01:23:55.883184Z","iopub.status.idle":"2024-12-09T01:23:55.890861Z","shell.execute_reply.started":"2024-12-09T01:23:55.883151Z","shell.execute_reply":"2024-12-09T01:23:55.889950Z"}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"# Train","metadata":{}},{"cell_type":"code","source":"from tensorflow.keras.callbacks import LearningRateScheduler\n\n# Define a learning rate scheduler function\ndef scheduler(epoch, lr):\n    if epoch > 25:\n        return lr * 0.5  # Halve the learning rate after 25 epochs\n    return lr\n\n# Attach the scheduler as a callback during training\nlr_scheduler_callback = LearningRateScheduler(scheduler)\n\nwith strategy.scope():\n    # Create generators\n    monet_generator_optimizer = optimizers.Adam(2e-4, beta_1=0.5)\n    photo_generator_optimizer = optimizers.Adam(2e-4, beta_1=0.5)\n\n    # Create discriminators\n    monet_discriminator_optimizer = optimizers.Adam(2e-4, beta_1=0.5)\n    photo_discriminator_optimizer = optimizers.Adam(2e-4, beta_1=0.5)\n\n    # Create GAN\n    gan_model = CycleGan(monet_generator, photo_generator, \n                         monet_discriminator, photo_discriminator)\n\n    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    )\n\n# Train with the learning rate scheduler as a callback\nhistory = gan_model.fit(\n    get_gan_dataset(MONET_FILENAMES, PHOTO_FILENAMES, batch_size=BATCH_SIZE), \n    steps_per_epoch=(n_monet_samples // BATCH_SIZE),\n    epochs=EPOCHS,\n    callbacks=[lr_scheduler_callback],\n    verbose=2\n)\n","metadata":{"trusted":true,"_kg_hide-output":true,"execution":{"iopub.status.busy":"2024-12-09T02:48:36.221366Z","iopub.execute_input":"2024-12-09T02:48:36.222324Z","iopub.status.idle":"2024-12-09T03:09:22.490378Z","shell.execute_reply.started":"2024-12-09T02:48:36.222266Z","shell.execute_reply":"2024-12-09T03:09:22.489588Z"}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"# Visualize predictions","metadata":{}},{"cell_type":"code","source":"display_generated_samples(load_dataset(PHOTO_FILENAMES,Is_Monet=False).batch(1), monet_generator, 8)","metadata":{"_kg_hide-input":true,"trusted":true,"execution":{"iopub.status.busy":"2024-12-09T01:46:04.484041Z","iopub.execute_input":"2024-12-09T01:46:04.484756Z","iopub.status.idle":"2024-12-09T01:47:16.493110Z","shell.execute_reply.started":"2024-12-09T01:46:04.484718Z","shell.execute_reply":"2024-12-09T01:47:16.492398Z"}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"## Make predictions","metadata":{}},{"cell_type":"code","source":"os.makedirs('../images/') # Create folder to save generated images\n\npredict_and_save(load_dataset(PHOTO_FILENAMES,Is_Monet=False).batch(1), monet_generator, '../images/')","metadata":{"trusted":true,"_kg_hide-input":true,"execution":{"iopub.status.busy":"2024-12-09T01:50:09.219105Z","iopub.execute_input":"2024-12-09T01:50:09.219841Z","iopub.status.idle":"2024-12-09T01:57:48.940249Z","shell.execute_reply.started":"2024-12-09T01:50:09.219807Z","shell.execute_reply":"2024-12-09T01:57:48.939473Z"}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"# Submission file","metadata":{}},{"cell_type":"code","source":"shutil.make_archive('/kaggle/working/images/', 'zip', '../images')\n\nprint(f\"Generated samples: {len([name for name in os.listdir('../images/') if os.path.isfile(os.path.join('../images/', name))])}\")","metadata":{"trusted":true,"_kg_hide-input":true,"execution":{"iopub.status.busy":"2024-12-09T02:00:33.696988Z","iopub.execute_input":"2024-12-09T02:00:33.697901Z","iopub.status.idle":"2024-12-09T02:00:37.253694Z","shell.execute_reply.started":"2024-12-09T02:00:33.697864Z","shell.execute_reply":"2024-12-09T02:00:37.252744Z"}},"outputs":[],"execution_count":null}]}