\n
\n\n### CycleGAN architecture\n\nLooking at the code below may be hard to get what is happening, this image will help the understanding\n\n
\n\nFirst, we get the regular generator discriminator thing, where the generator tries to generate images that seem to be drawn to the given domain (in the example will be creating zebra images), but it would be possible that the generator generates only the same zebra image or zebra images that do not look like the imputed horse image, this is why the model has a second generator, this second generator uses the first generated image and tries to recreate the original imputed horse image, this way the first generator has to generate zebra images that look like the imputed horse image.\n\n#### In the end, you will get 4 sub-models:\n- A generator that can generate zebras images\n- A generator that can generate horses images\n- A discriminator that can identify real zebras images\n- A discriminator that can identify real horses images\n\n\n#### Let's get to the code","metadata":{}},{"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":"2025-05-22T15:42:41.338097Z","iopub.execute_input":"2025-05-22T15:42:41.338856Z","iopub.status.idle":"2025-05-22T15:42:47.921085Z","shell.execute_reply.started":"2025-05-22T15:42:41.338774Z","shell.execute_reply":"2025-05-22T15:42:47.920107Z"}},"outputs":[],"execution_count":1},{"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":"2025-05-22T15:43:22.59902Z","iopub.execute_input":"2025-05-22T15:43:22.600004Z","iopub.status.idle":"2025-05-22T15:43:22.613077Z","shell.execute_reply.started":"2025-05-22T15:43:22.599969Z","shell.execute_reply":"2025-05-22T15:43:22.612076Z"}},"outputs":[{"name":"stdout","text":"REPLICAS: 1\n","output_type":"stream"}],"execution_count":2},{"cell_type":"markdown","source":"# Model parameters","metadata":{}},{"cell_type":"code","source":"HEIGHT = 256\nWIDTH = 256\nCHANNELS = 3\nEPOCHS = 50\nBATCH_SIZE = 1","metadata":{"_kg_hide-input":false,"trusted":true,"execution":{"iopub.status.busy":"2025-05-22T15:43:25.628509Z","iopub.execute_input":"2025-05-22T15:43:25.629341Z","iopub.status.idle":"2025-05-22T15:43:25.633348Z","shell.execute_reply.started":"2025-05-22T15:43:25.629309Z","shell.execute_reply":"2025-05-22T15:43:25.632428Z"}},"outputs":[],"execution_count":3},{"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)\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":"2025-05-22T15:43:30.438565Z","iopub.execute_input":"2025-05-22T15:43:30.439265Z","iopub.status.idle":"2025-05-22T15:43:31.240555Z","shell.execute_reply.started":"2025-05-22T15:43:30.439203Z","shell.execute_reply":"2025-05-22T15:43:31.239623Z"}},"outputs":[{"name":"stdout","text":"Monet TFRecord files: 5\nMonet image files: 300\nPhoto TFRecord files: 20\nPhoto image files: 7038\n","output_type":"stream"}],"execution_count":4},{"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 load_dataset(filenames):\n dataset = tf.data.TFRecordDataset(filenames)\n dataset = dataset.map(read_tfrecord, num_parallel_calls=AUTO)\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)\n photo_ds = load_dataset(photo_files)\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 \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":"2025-05-22T15:43:36.283977Z","iopub.execute_input":"2025-05-22T15:43:36.284332Z","iopub.status.idle":"2025-05-22T15:43:36.302265Z","shell.execute_reply.started":"2025-05-22T15:43:36.284303Z","shell.execute_reply":"2025-05-22T15:43:36.301387Z"}},"outputs":[],"execution_count":5},{"cell_type":"markdown","source":"## Looking at a few Monet paintings","metadata":{}},{"cell_type":"code","source":"display_samples(load_dataset(MONET_FILENAMES).batch(1), 4, 6)","metadata":{"_kg_hide-input":true,"trusted":true,"execution":{"iopub.status.busy":"2025-05-22T15:43:44.943759Z","iopub.execute_input":"2025-05-22T15:43:44.944438Z","iopub.status.idle":"2025-05-22T15:43:50.289539Z","shell.execute_reply.started":"2025-05-22T15:43:44.944406Z","shell.execute_reply":"2025-05-22T15:43:50.288641Z"}},"outputs":[{"output_type":"display_data","data":{"text/plain":"