{"cells":[{"metadata":{"_cell_guid":"dbc2a207-a9a7-4021-83c6-d6756c328376","_uuid":"016e823c-0640-498b-b542-c7dbee91f0f8"},"cell_type":"markdown","source":"#  Flower Classification with TPUs - EfficientNetB7 + DenseNet-201 Soft Voting & Mixup\nNothing special, just for self-study\n\n**Notes**\n\n  * Used the [Getting Started notebook](https://www.kaggle.com/mgornergoogle/getting-started-with-100-flowers-on-tpu/) for this competition as a base.\n\n  * Added a few modifications to enable it to run on my local machine. (e.g. `my_tf_initialize()` )\n\n  * Then implemented EfficientNetB7 + DenseNet-201 Soft Voting and a [learning rate scheduler](https://codelabs.developers.google.com/codelabs/keras-flowers-tpu/#11)\n  \n  * Added mixup (mixup: Beyond Empirical Risk Minimization, Hongyi Zhang, Moustapha Cisse, Yann N. Dauphin, David Lopez-Paz, arXiv preprint [arXiv:1710.09412](https://arxiv.org/abs/1710.09412))"},{"metadata":{"trusted":false},"cell_type":"code","source":"!pip install -q efficientnet","execution_count":null,"outputs":[]},{"metadata":{"_cell_guid":"73d408a9-67ef-4590-a4ee-55f612766226","_uuid":"7a95c36d-2def-4de7-a61a-c31c95f6bc2f","scrolled":false,"trusted":false},"cell_type":"code","source":"# Flower Classification with TPUs \n# Copied from the Getting Started Notebook for this competition\n# https://www.kaggle.com/mgornergoogle/getting-started-with-100-flowers-on-tpu/\n\nimport math, re, os\nimport tensorflow as tf\nimport numpy as np\nfrom matplotlib import pyplot as plt\nfrom sklearn.metrics import f1_score, precision_score, recall_score, confusion_matrix\n\n# My imports\nimport json\nimport seaborn as sns\nfrom datetime import datetime\nstart_time = datetime.now()\nprint('start time:', start_time)\n\nprint(\"Tensorflow version \" + tf.__version__)\nAUTO = tf.data.experimental.AUTOTUNE\n\ndef my_tf_initialize():\n    gpus = tf.config.experimental.list_physical_devices('GPU')\n    if gpus:\n        try:\n            for gpu in gpus:\n                tf.config.experimental.set_memory_growth(gpu, True)\n                logical_gpus = tf.config.experimental.list_logical_devices('GPU')\n                print(len(gpus), \"Physical GPUs,\", len(logical_gpus), \"Logical GPUs\")\n        except RuntimeError as e:\n            print(e)\n    tf.keras.backend.clear_session()\n    # make the notebook's output stable across runs\n    np.random.seed(42)\n    tf.random.set_seed(42)\n\n# Detect hardware, return appropriate distribution strategy\ntry:\n    tpu = tf.distribute.cluster_resolver.TPUClusterResolver()  # TPU detection. No parameters necessary if TPU_NAME environment variable is set. On Kaggle this is always the case.\n    print('Running on TPU ', tpu.master())\nexcept ValueError:\n    tpu = None\n    my_tf_initialize() # required for my machine\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() # default distribution strategy in Tensorflow. Works on CPU and single GPU.\n\nprint(\"REPLICAS: \", strategy.num_replicas_in_sync)\n\nMIXED_PRECISION = True\nif MIXED_PRECISION:\n    if tpu: \n        # Disable due to errors on Kaggle's TPU Accelerator\n        # policy = tf.keras.mixed_precision.experimental.Policy('mixed_bfloat16')\n        # policy = tf.keras.mixed_precision.experimental.Policy('mixed_float16')\n        print('Mixed precision not enabled')\n    else:\n        policy = tf.keras.mixed_precision.experimental.Policy('mixed_float16')\n        tf.config.optimizer.set_jit(True) # XLA compilation\n        tf.keras.mixed_precision.experimental.set_policy(policy)\n        print('Mixed precision enabled')\n","execution_count":null,"outputs":[]},{"metadata":{"_cell_guid":"a5a2a15c-972a-4ba8-9e9b-797ee52a6114","_uuid":"db9f151c-eec9-48d0-a904-2307fcdc7455","scrolled":false,"trusted":false},"cell_type":"code","source":"# TPUs read data directly from Google Cloud Storage (GCS). This Kaggle utility will copy the dataset to \n# a GCS bucket co-located with the TPU. If you have multiple datasets attached to the notebook, you can \n# pass the name of a specific dataset to the get_gcs_path function. The name of the dataset is the name \n# of the directory it is mounted in. Use !ls /kaggle/input/ to list attached datasets.\n\nif tpu:\n    from kaggle_datasets import KaggleDatasets\n    GCS_DS_PATH = KaggleDatasets().get_gcs_path('flower-classification-with-tpus')\nelse:\n    GCS_DS_PATH = os.getcwd() # local machine\n    print(GCS_DS_PATH)","execution_count":null,"outputs":[]},{"metadata":{"_cell_guid":"05cf18d5-6f65-48d0-ba1c-2fe6fa6fa53a","_uuid":"f740651b-1ca7-4051-9774-3b140589b88d","scrolled":false,"trusted":false},"cell_type":"code","source":"# Configuration\nif tpu:\n    IMAGE_SIZE = [512, 512] # At this size, a GPU will run out of memory. Use the TPU.\n                            # For GPU training, please select 224 x 224 px image size.\nelse:\n    #IMAGE_SIZE = [224, 224] # At this size, a GPU will run out of memory. Use the TPU.\n    IMAGE_SIZE = [192, 192] # At this size, a GPU will run out of memory. Use the TPU.\n                            # For GPU training, please select 224 x 224 px image size.\n        \nEPOCHS = 20\nBATCH_SIZE = 16 * strategy.num_replicas_in_sync\nif tpu:\n    BATCH_SIZE = 16 * strategy.num_replicas_in_sync\nelse:\n    BATCH_SIZE = 16 * strategy.num_replicas_in_sync\n    BATCH_SIZE = 4 * strategy.num_replicas_in_sync\n\nGCS_PATH_SELECT = { # available image sizes\n    192: GCS_DS_PATH + '/tfrecords-jpeg-192x192',\n    224: GCS_DS_PATH + '/tfrecords-jpeg-224x224',\n    331: GCS_DS_PATH + '/tfrecords-jpeg-331x331',\n    512: GCS_DS_PATH + '/tfrecords-jpeg-512x512'\n}\nGCS_PATH = GCS_PATH_SELECT[IMAGE_SIZE[0]]\n\nTRAINING_FILENAMES = tf.io.gfile.glob(GCS_PATH + '/train/*.tfrec')\nVALIDATION_FILENAMES = tf.io.gfile.glob(GCS_PATH + '/val/*.tfrec')\nTEST_FILENAMES = tf.io.gfile.glob(GCS_PATH + '/test/*.tfrec') # predictions on this dataset should be submitted for the competition\n\nCLASSES = ['pink primrose',    'hard-leaved pocket orchid', 'canterbury bells', 'sweet pea',     'wild geranium',     'tiger lily',           'moon orchid',              'bird of paradise', 'monkshood',        'globe thistle',         # 00 - 09\n           'snapdragon',       \"colt's foot\",               'king protea',      'spear thistle', 'yellow iris',       'globe-flower',         'purple coneflower',        'peruvian lily',    'balloon flower',   'giant white arum lily', # 10 - 19\n           'fire lily',        'pincushion flower',         'fritillary',       'red ginger',    'grape hyacinth',    'corn poppy',           'prince of wales feathers', 'stemless gentian', 'artichoke',        'sweet william',         # 20 - 29\n           'carnation',        'garden phlox',              'love in the mist', 'cosmos',        'alpine sea holly',  'ruby-lipped cattleya', 'cape flower',              'great masterwort', 'siam tulip',       'lenten rose',           # 30 - 39\n           'barberton daisy',  'daffodil',                  'sword lily',       'poinsettia',    'bolero deep blue',  'wallflower',           'marigold',                 'buttercup',        'daisy',            'common dandelion',      # 40 - 49\n           'petunia',          'wild pansy',                'primula',          'sunflower',     'lilac hibiscus',    'bishop of llandaff',   'gaura',                    'geranium',         'orange dahlia',    'pink-yellow dahlia',    # 50 - 59\n           'cautleya spicata', 'japanese anemone',          'black-eyed susan', 'silverbush',    'californian poppy', 'osteospermum',         'spring crocus',            'iris',             'windflower',       'tree poppy',            # 60 - 69\n           'gazania',          'azalea',                    'water lily',       'rose',          'thorn apple',       'morning glory',        'passion flower',           'lotus',            'toad lily',        'anthurium',             # 70 - 79\n           'frangipani',       'clematis',                  'hibiscus',         'columbine',     'desert-rose',       'tree mallow',          'magnolia',                 'cyclamen ',        'watercress',       'canna lily',            # 80 - 89\n           'hippeastrum ',     'bee balm',                  'pink quill',       'foxglove',      'bougainvillea',     'camellia',             'mallow',                   'mexican petunia',  'bromelia',         'blanket flower',        # 90 - 99\n           'trumpet creeper',  'blackberry lily',           'common tulip',     'wild rose']                                                                                                                                               # 100 - 102","execution_count":null,"outputs":[]},{"metadata":{"_cell_guid":"d4315e58-4a13-470f-9296-14a026f1b47e","_uuid":"49941dc7-0266-4ddc-a68d-c2e6fc2f3e36","scrolled":false,"trusted":false},"cell_type":"code","source":"# Visualization utilities\n# numpy and matplotlib defaults\nnp.set_printoptions(threshold=15, linewidth=80)\n\ndef batch_to_numpy_images_and_labels(data):\n    images, labels = data\n    numpy_images = images.numpy()\n    numpy_labels = labels.numpy()\n    if numpy_labels.dtype == object: # binary string in this case, these are image ID strings\n        numpy_labels = [None for _ in enumerate(numpy_images)]\n    # If no labels, only image IDs, return None for labels (this is the case for test data)\n    return numpy_images, numpy_labels\n\ndef title_from_label_and_target(label, correct_label):\n    if correct_label is None:\n        #return CLASSES[label], True\n        return CLASSES[np.argmax(label, axis=-1)], True\n    correct = (label == correct_label)\n    #return \"{} [{}{}{}]\".format(CLASSES[label], 'OK' if correct else 'NO', u\"\\u2192\" if not correct else '',\n    #                            CLASSES[correct_label] if not correct else ''), correct\n    return \"{} [{}{}{}]\".format(CLASSES[np.argmax(label, axis=-1)], 'OK' if correct else 'NO', u\"\\u2192\" if not correct else '',\n                                CLASSES[np.argmax(correct_label, axis=-1)] if not correct else ''), correct\n\ndef display_one_flower(image, title, subplot, red=False, titlesize=16):\n    plt.subplot(*subplot)\n    plt.axis('off')\n    plt.imshow(image)\n    if len(title) > 0:\n        plt.title(title, fontsize=int(titlesize) if not red else int(titlesize/1.2), color='red' if red else 'black', fontdict={'verticalalignment':'center'}, pad=int(titlesize/1.5))\n    return (subplot[0], subplot[1], subplot[2]+1)\n    \ndef display_batch_of_images(databatch, predictions=None):\n    \"\"\"This will work with:\n    display_batch_of_images(images)\n    display_batch_of_images(images, predictions)\n    display_batch_of_images((images, labels))\n    display_batch_of_images((images, labels), predictions)\n    \"\"\"\n    # data\n    images, labels = batch_to_numpy_images_and_labels(databatch)\n    if labels is None:\n        labels = [None for _ in enumerate(images)]\n        \n    # auto-squaring: this will drop data that does not fit into square or square-ish rectangle\n    rows = int(math.sqrt(len(images)))\n    cols = len(images)//rows\n        \n    # size and spacing\n    FIGSIZE = 13.0\n    SPACING = 0.1\n    subplot=(rows,cols,1)\n    if rows < cols:\n        plt.figure(figsize=(FIGSIZE,FIGSIZE/cols*rows))\n    else:\n        plt.figure(figsize=(FIGSIZE/rows*cols,FIGSIZE))\n    \n    # display\n    for i, (image, label) in enumerate(zip(images[:rows*cols], labels[:rows*cols])):\n        #title = '' if label is None else CLASSES[label]\n        title = '' if label is None else CLASSES[np.argmax(label, axis=-1)]\n        correct = True\n        if predictions is not None:\n            title, correct = title_from_label_and_target(predictions[i], label)\n        dynamic_titlesize = FIGSIZE*SPACING/max(rows,cols)*40+3 # magic formula tested to work from 1x1 to 10x10 images\n        subplot = display_one_flower(image, title, subplot, not correct, titlesize=dynamic_titlesize)\n    \n    #layout\n    plt.tight_layout()\n    if label is None and predictions is None:\n        plt.subplots_adjust(wspace=0, hspace=0)\n    else:\n        plt.subplots_adjust(wspace=SPACING, hspace=SPACING)\n    plt.show()\n\ndef display_confusion_matrix(cmat, score, precision, recall):\n    plt.figure(figsize=(15,15))\n    ax = plt.gca()\n    ax.matshow(cmat, cmap='Reds')\n    ax.set_xticks(range(len(CLASSES)))\n    ax.set_xticklabels(CLASSES, fontdict={'fontsize': 7})\n    plt.setp(ax.get_xticklabels(), rotation=45, ha=\"left\", rotation_mode=\"anchor\")\n    ax.set_yticks(range(len(CLASSES)))\n    ax.set_yticklabels(CLASSES, fontdict={'fontsize': 7})\n    plt.setp(ax.get_yticklabels(), rotation=45, ha=\"right\", rotation_mode=\"anchor\")\n    titlestring = \"\"\n    if score is not None:\n        titlestring += 'f1 = {:.3f} '.format(score)\n    if precision is not None:\n        titlestring += '\\nprecision = {:.3f} '.format(precision)\n    if recall is not None:\n        titlestring += '\\nrecall = {:.3f} '.format(recall)\n    if len(titlestring) > 0:\n        ax.text(101, 1, titlestring, fontdict={'fontsize': 18, 'horizontalalignment':'right', 'verticalalignment':'top', 'color':'#804040'})\n    plt.show()\n    \ndef display_training_curves(training, validation, title, subplot):\n    if subplot%10==1: # set up the subplots on the first call\n        plt.subplots(figsize=(10,10), facecolor='#F0F0F0')\n        plt.tight_layout()\n    ax = plt.subplot(subplot)\n    ax.set_facecolor('#F8F8F8')\n    ax.plot(training)\n    ax.plot(validation)\n    ax.set_title('model '+ title)\n    ax.set_ylabel(title)\n    #ax.set_ylim(0.28,1.05)\n    ax.set_xlabel('epoch')\n    ax.legend(['train', 'valid.'])","execution_count":null,"outputs":[]},{"metadata":{"_cell_guid":"2422d201-465a-4387-9d22-6a0c9ee221c2","_uuid":"88bd3e20-e919-42bd-b149-f2cf6400fa53","scrolled":false,"trusted":false},"cell_type":"code","source":"# Datasets\ndef decode_image(image_data):\n    image = tf.image.decode_jpeg(image_data, channels=3)\n    image = tf.cast(image, tf.float32) / 255.0  # convert image to floats in [0, 1] range\n    image = tf.reshape(image, [*IMAGE_SIZE, 3]) # explicit size needed for TPU\n    return image\n\ndef read_labeled_tfrecord(example):\n    LABELED_TFREC_FORMAT = {\n        \"image\": tf.io.FixedLenFeature([], tf.string), # tf.string means bytestring\n        \"class\": tf.io.FixedLenFeature([], tf.int64),  # shape [] means single element\n    }\n    example = tf.io.parse_single_example(example, LABELED_TFREC_FORMAT)\n    image = decode_image(example['image'])\n    label = tf.cast(example['class'], tf.int32)\n    return image, label # returns a dataset of (image, label) pairs\n\ndef read_unlabeled_tfrecord(example):\n    UNLABELED_TFREC_FORMAT = {\n        \"image\": tf.io.FixedLenFeature([], tf.string), # tf.string means bytestring\n        \"id\": tf.io.FixedLenFeature([], tf.string),  # shape [] means single element\n        # class is missing, this competitions's challenge is to predict flower classes for the test dataset\n    }\n    example = tf.io.parse_single_example(example, UNLABELED_TFREC_FORMAT)\n    image = decode_image(example['image'])\n    idnum = example['id']\n    return image, idnum # returns a dataset of image(s)\n\ndef load_dataset(filenames, labeled=True, ordered=False):\n    # Read from TFRecords. For optimal performance, reading from multiple files at once and\n    # disregarding data order. Order does not matter since we will be shuffling the data anyway.\n\n    ignore_order = tf.data.Options()\n    if not ordered:\n        ignore_order.experimental_deterministic = False # disable order, increase speed\n\n    dataset = tf.data.TFRecordDataset(filenames, num_parallel_reads=AUTO) # automatically interleaves reads from multiple files\n    dataset = dataset.with_options(ignore_order) # uses data as soon as it streams in, rather than in its original order\n    dataset = dataset.map(read_labeled_tfrecord if labeled else read_unlabeled_tfrecord, num_parallel_calls=AUTO)\n    # returns a dataset of (image, label) pairs if labeled=True or (image, id) pairs if labeled=False\n    return dataset\n\ndef data_to_label(image, label):\n    return label\n\ndef data_to_categorical(image, label):\n    #label = tf.keras.utils.to_categorical(label, len(CLASSES))\n    label = tf.one_hot(label, depth=len(CLASSES))\n    return image, label   \n\ndef data_augment(image, label):\n    # data augmentation. Thanks to the dataset.prefetch(AUTO) statement in the next function (below),\n    # this happens essentially for free on TPU. Data pipeline code is executed on the \"CPU\" part\n    # of the TPU while the TPU itself is computing gradients.\n    image = tf.image.random_flip_left_right(image)\n    image = tf.image.random_saturation(image, 0, 2)\n    return image, label   \n\nimport random\n\ndef element_mixup(a, b):\n    X_a, y_a = a\n    X_b, y_b = b\n    lambda_1 = tf.random.uniform(shape=[], minval=0., maxval=1.)\n    lambda_2 = tf.math.subtract(tf.constant(1.), lambda_1)\n    X = tf.math.add(tf.math.scalar_mul(lambda_1, X_a), tf.math.scalar_mul(lambda_2, X_b))\n    y = tf.math.add(tf.math.scalar_mul(lambda_1, y_a), tf.math.scalar_mul(lambda_2, y_b))\n    #if lambda_1 > 0.5:\n    #    y = y_a\n    #else:\n    #    y = y_b\n    return tf.data.Dataset.from_tensors((X, y))\n    \n#def dataset_mixup(dataset_a):\ndef dataset_mixup(dataset_a):\n    dataset_b = dataset_a.shuffle(2048)\n    dataset = tf.data.Dataset.zip((dataset_a, dataset_b))\n    dataset = dataset.flat_map(lambda a, b: element_mixup(a, b))\n    return dataset   \n\ndef get_training_dataset():\n    dataset = load_dataset(TRAINING_FILENAMES, labeled=True)\n    dataset = dataset.map(data_to_categorical, num_parallel_calls=AUTO)\n    dataset = dataset.map(data_augment, num_parallel_calls=AUTO)\n    #dataset = dataset_mixup(dataset)\n    dataset = dataset.repeat() # the training dataset must repeat for several epochs\n    dataset = dataset.shuffle(2048)\n    dataset = dataset.batch(BATCH_SIZE)\n    dataset = dataset.prefetch(AUTO) # prefetch next batch while training (autotune prefetch buffer size)\n    return dataset\n\ndef get_training_dataset_mixup():\n    dataset = load_dataset(TRAINING_FILENAMES, labeled=True)\n    dataset = dataset.map(data_to_categorical, num_parallel_calls=AUTO)\n    #dataset = dataset.map(data_augment, num_parallel_calls=AUTO)\n    dataset = dataset_mixup(dataset)\n    dataset = dataset.repeat() # the training dataset must repeat for several epochs\n    dataset = dataset.shuffle(2048)\n    dataset = dataset.batch(BATCH_SIZE)\n    dataset = dataset.prefetch(AUTO) # prefetch next batch while training (autotune prefetch buffer size)\n    return dataset\n\ndef get_validation_dataset_labels(ordered=False):\n    dataset = load_dataset(VALIDATION_FILENAMES, labeled=True, ordered=ordered)\n    dataset = dataset.map(data_to_label, num_parallel_calls=AUTO)\n    #dataset = dataset.batch(BATCH_SIZE)\n    dataset = dataset.cache()\n    dataset = dataset.prefetch(AUTO) # prefetch next batch while training (autotune prefetch buffer size)\n    return dataset\n\ndef get_validation_dataset(ordered=False):\n    dataset = load_dataset(VALIDATION_FILENAMES, labeled=True, ordered=ordered)\n    dataset = dataset.map(data_to_categorical, num_parallel_calls=AUTO)\n    dataset = dataset.batch(BATCH_SIZE)\n    dataset = dataset.cache()\n    dataset = dataset.prefetch(AUTO) # prefetch next batch while training (autotune prefetch buffer size)\n    return dataset\n\ndef get_test_dataset(ordered=False):\n    dataset = load_dataset(TEST_FILENAMES, labeled=False, ordered=ordered)\n    dataset = dataset.batch(BATCH_SIZE)\n    dataset = dataset.prefetch(AUTO) # prefetch next batch while training (autotune prefetch buffer size)\n    return dataset\n\ndef count_data_items(filenames):\n    # the number of data items is written in the name of the .tfrec files, i.e. flowers00-230.tfrec = 230 data items\n    n = [int(re.compile(r\"-([0-9]*)\\.\").search(filename).group(1)) for filename in filenames]\n    return np.sum(n)\n\nNUM_TRAINING_IMAGES = count_data_items(TRAINING_FILENAMES)\nNUM_VALIDATION_IMAGES = count_data_items(VALIDATION_FILENAMES)\nNUM_TEST_IMAGES = count_data_items(TEST_FILENAMES)\nSTEPS_PER_EPOCH = NUM_TRAINING_IMAGES // BATCH_SIZE\nprint('Dataset: {} training images, {} validation images, {} unlabeled test images'.format(NUM_TRAINING_IMAGES, NUM_VALIDATION_IMAGES, NUM_TEST_IMAGES))\n\n# Dataset visualizations\n\n# data dump\nprint(\"Training data shapes:\")\nfor image, label in get_training_dataset().take(3):\n    print(image.numpy().shape, label.numpy().shape)\nprint(\"Training data label examples:\", label.numpy())\nprint(\"Validation data shapes:\")\nfor image, label in get_validation_dataset().take(3):\n    print(image.numpy().shape, label.numpy().shape)\nprint(\"Validation data label examples:\", label.numpy())\nprint(\"Test data shapes:\")\nfor image, idnum in get_test_dataset().take(3):\n    print(image.numpy().shape, idnum.numpy().shape)\nprint(\"Test data IDs:\", idnum.numpy().astype('U')) # U=unicode string\n\n# Peek at training data\ntraining_dataset = get_training_dataset()\ntraining_dataset = training_dataset.unbatch().batch(20)\ntrain_batch = iter(training_dataset)\n\n# peer at test data\ntest_dataset = get_test_dataset()\ntest_dataset = test_dataset.unbatch().batch(20)\ntest_batch = iter(test_dataset)\n","execution_count":null,"outputs":[]},{"metadata":{"_cell_guid":"a85ebf82-0365-4edd-bcd0-301767ab0e44","_uuid":"5154094b-dc9d-4170-9e45-7347f30d4749","trusted":false},"cell_type":"code","source":"from sklearn.utils import class_weight\nlabels = get_validation_dataset_labels().take(NUM_VALIDATION_IMAGES)\nlabels = list(labels.as_numpy_iterator())\nsns.countplot(labels)\nclass_weights = class_weight.compute_class_weight('balanced', np.unique(labels), labels)\n#%xdel labels","execution_count":null,"outputs":[]},{"metadata":{"_cell_guid":"e38635d5-22bf-41af-a15e-ff4c20d83f52","_uuid":"2077cab2-852a-4a56-b902-1d912980aabd","scrolled":false,"trusted":false},"cell_type":"code","source":"# run this cell again for next set of images\ndisplay_batch_of_images(next(train_batch))","execution_count":null,"outputs":[]},{"metadata":{"_cell_guid":"1b69c8c7-9ab9-415e-b6c0-8c44277b2018","_uuid":"fa15018d-6cc4-40e1-a8f4-9a242b3fca2c","scrolled":false,"trusted":false},"cell_type":"code","source":"# run this cell again for next set of images\ndisplay_batch_of_images(next(test_batch))","execution_count":null,"outputs":[]},{"metadata":{"trusted":false},"cell_type":"code","source":"early_stopping_cb = tf.keras.callbacks.EarlyStopping(monitor='val_loss', min_delta=0.01, patience=5, restore_best_weights=True)\n\nif tpu:\n    start_lr = 0.00001\n    min_lr = 0.00001\n    max_lr = 0.00005 * strategy.num_replicas_in_sync\n    rampup_epochs = 5\n    sustain_epochs = 0\n    exp_decay = .8\nelse:\n    start_lr = 0.00001\n    min_lr = 0.00001\n    max_lr = 0.0002\n    rampup_epochs = 5\n    sustain_epochs = 0\n    exp_decay = .8\n\ndef lrfn(epoch):\n    def lr(epoch, start_lr, min_lr, max_lr, rampup_epochs, sustain_epochs, exp_decay):\n        if epoch < rampup_epochs:\n            lr = (max_lr - start_lr)/rampup_epochs * epoch + start_lr\n        elif epoch < rampup_epochs + sustain_epochs:\n            lr = max_lr\n        else:\n            lr = (max_lr - min_lr) * exp_decay**(epoch-rampup_epochs-sustain_epochs) + min_lr\n        return lr\n    return lr(epoch, start_lr, min_lr, max_lr, rampup_epochs, sustain_epochs, exp_decay)\n    \nlr_callback = tf.keras.callbacks.LearningRateScheduler(lambda epoch: lrfn(epoch), verbose=True)\n\nrng = [i for i in range(EPOCHS)]\ny = [lrfn(x) for x in rng]\nplt.plot(rng, [lrfn(x) for x in rng])\nprint(y[0], y[-1])","execution_count":null,"outputs":[]},{"metadata":{"_cell_guid":"d8d7c35d-1609-44b7-9089-dfbb6658a502","_uuid":"41f5fabd-6d9b-4f9e-b323-071e9f2ecb78","scrolled":true,"trusted":false},"cell_type":"code","source":"# Models\nimport efficientnet.tfkeras as efntf\nfrom functools import partial\n\nwith strategy.scope():\n    DefaultConv2D = partial(tf.keras.layers.Conv2D,\n                        kernel_size=3, activation='relu', padding=\"SAME\")\n    \n    simple_cnn_model = tf.keras.models.Sequential([\n        DefaultConv2D(filters=64, kernel_size=7, input_shape=[*IMAGE_SIZE, 3]),\n        tf.keras.layers.MaxPooling2D(pool_size=2),\n        DefaultConv2D(filters=128),\n        DefaultConv2D(filters=128),\n        tf.keras.layers.MaxPooling2D(pool_size=2),\n        DefaultConv2D(filters=256),\n        DefaultConv2D(filters=256),\n        tf.keras.layers.GlobalAveragePooling2D(),\n        tf.keras.layers.Dense(len(CLASSES), activation=\"softmax\")\n        #tf.keras.layers.Activation(activation=\"softmax\")\n    ])\n    \n    simple_cnn_model.compile(\n        optimizer='adam',\n        loss = 'categorical_crossentropy',\n        metrics=['categorical_accuracy']\n    )\n    #simple_cnn_model.summary()\n        \n    ##base_model = tf.keras.applications.densenet.DenseNet121(weights='imagenet', include_top=False ,input_shape=[*IMAGE_SIZE, 3])\n    ##base_model = tf.keras.applications.densenet.DenseNet169(weights='imagenet', include_top=False ,input_shape=[*IMAGE_SIZE, 3])\n    #base_model = tf.keras.applications.densenet.DenseNet201(weights='imagenet', include_top=False ,input_shape=[*IMAGE_SIZE, 3])\n    #base_model = tf.keras.applications.inception_resnet_v2.InceptionResNetV2(weights='imagenet', include_top=False ,input_shape=[*IMAGE_SIZE, 3])\n    #base_model = tf.keras.applications.inception_v3.InceptionV3(weights='imagenet', include_top=False ,input_shape=[*IMAGE_SIZE, 3])\n    ##base_model = tf.keras.applications.mobilenet.MobileNet(weights='imagenet', include_top=False ,input_shape=[*IMAGE_SIZE, 3])\n    #base_model = tf.keras.applications.mobilenet_v2.MobileNetV2(weights='imagenet', include_top=False ,input_shape=[*IMAGE_SIZE, 3])\n    ##base_model = tf.keras.applications.resnet50.ResNet50(weights='imagenet', include_top=False ,input_shape=[*IMAGE_SIZE, 3])\n    ##base_model = tf.keras.applications.resnet.ResNet101(weights='imagenet', include_top=False ,input_shape=[*IMAGE_SIZE, 3])\n    ##base_model = tf.keras.applications.resnet.ResNet152(weights='imagenet', include_top=False ,input_shape=[*IMAGE_SIZE, 3])\n    ##base_model = tf.keras.applications.resnet_v2.ResNet101V2(weights='imagenet', include_top=False ,input_shape=[*IMAGE_SIZE, 3])\n    #base_model = tf.keras.applications.resnet_v2.ResNet152V2(weights='imagenet', include_top=False ,input_shape=[*IMAGE_SIZE, 3])\n    #base_model = tf.keras.applications.resnet_v2.ResNet50V2(weights='imagenet', include_top=False ,input_shape=[*IMAGE_SIZE, 3])\n    #base_model = tf.keras.applications.vgg16.VGG16(weights='imagenet', include_top=False ,input_shape=[*IMAGE_SIZE, 3])\n    #base_model = tf.keras.applications.vgg19.VGG19(weights='imagenet', include_top=False ,input_shape=[*IMAGE_SIZE, 3])\n    #base_model = tf.keras.applications.xception.Xception(weights='imagenet', include_top=False ,input_shape=[*IMAGE_SIZE, 3])\n    \n    enetb7_base_model = efntf.EfficientNetB7(weights='imagenet', include_top=False ,input_shape=[*IMAGE_SIZE, 3])\n    enetb7_avg = tf.keras.layers.GlobalAveragePooling2D()(enetb7_base_model.output)\n    enetb7_output = tf.keras.layers.Dense(len(CLASSES), activation=\"softmax\")(enetb7_avg)\n    enetb7_model = tf.keras.Model(inputs=enetb7_base_model.input, outputs=enetb7_output)\n    enetb7_model.trainable = True\n    \n    enetb7_base_model2 = efntf.EfficientNetB7(weights='imagenet', include_top=False ,input_shape=[*IMAGE_SIZE, 3])\n    enetb7_avg2 = tf.keras.layers.GlobalAveragePooling2D()(enetb7_base_model2.output)\n    enetb7_output2 = tf.keras.layers.Dense(len(CLASSES), activation=\"softmax\")(enetb7_avg2)\n    enetb7_model2 = tf.keras.Model(inputs=enetb7_base_model2.input, outputs=enetb7_output2)\n    enetb7_model2.trainable = True\n    \n    dense201_base_model = tf.keras.applications.densenet.DenseNet201(weights='imagenet', include_top=False ,input_shape=[*IMAGE_SIZE, 3])\n    dense201_avg = tf.keras.layers.GlobalAveragePooling2D()(dense201_base_model.output)\n    dense201_output = tf.keras.layers.Dense(len(CLASSES), activation=\"softmax\")(dense201_avg)\n    dense201_model = tf.keras.Model(inputs=dense201_base_model.input, outputs=dense201_output)\n    dense201_model.trainable = True\n    \n    dense201_base_model2 = tf.keras.applications.densenet.DenseNet201(weights='imagenet', include_top=False ,input_shape=[*IMAGE_SIZE, 3])\n    dense201_avg2 = tf.keras.layers.GlobalAveragePooling2D()(dense201_base_model2.output)\n    dense201_output2 = tf.keras.layers.Dense(len(CLASSES), activation=\"softmax\")(dense201_avg2)\n    dense201_model2 = tf.keras.Model(inputs=dense201_base_model2.input, outputs=dense201_output2)\n    dense201_model2.trainable = True\n    \n    enetb7_model.compile(\n        optimizer='adam',\n        loss = 'categorical_crossentropy',\n        metrics=['categorical_accuracy']\n    )\n    \n    enetb7_model2.compile(\n        optimizer='adam',\n        loss = 'categorical_crossentropy',\n        metrics=['categorical_accuracy']\n    )\n    \n    dense201_model.compile(\n        optimizer='adam',\n        loss = 'categorical_crossentropy',\n        metrics=['categorical_accuracy']\n    )\n\n    dense201_model2.compile(\n        optimizer='adam',\n        loss = 'categorical_crossentropy',\n        metrics=['categorical_accuracy']\n    )\n\n    #model.summary()\n","execution_count":null,"outputs":[]},{"metadata":{"_cell_guid":"81ca92ca-ec06-44c7-a73b-8d58de8579d1","_uuid":"ab2f47bf-0543-4f66-8e8f-09ea32fd3a1c","scrolled":false,"trusted":false},"cell_type":"code","source":"model = enetb7_model\n#model = enetb7_model2\n#model = xcept_model\n#model = dense201_model\n\nhistory = model.fit(get_training_dataset(), \n                    steps_per_epoch=STEPS_PER_EPOCH, epochs=EPOCHS,\n                    validation_data=get_validation_dataset(),\n                    callbacks=[early_stopping_cb, lr_callback])\n \ndisplay_training_curves(history.history['loss'], history.history['val_loss'], 'loss', 211)\ndisplay_training_curves(history.history['categorical_accuracy'], history.history['val_categorical_accuracy'], 'accuracy', 212)\n\n# Confusion matrix\ncmdataset = get_validation_dataset(ordered=True) # since we are splitting the dataset and iterating separately on images and labels, order matters.\nimages_ds = cmdataset.map(lambda image, label: image)\nlabels_ds = cmdataset.map(lambda image, label: label).unbatch()\ncm_correct_labels = next(iter(labels_ds.batch(NUM_VALIDATION_IMAGES))).numpy() # get everything as one batch\ncm_correct_labels = np.argmax(cm_correct_labels, axis=-1)\ncm_probabilities = model.predict(images_ds)\ncm_predictions = np.argmax(cm_probabilities, axis=-1)\nprint(\"Correct   labels: \", cm_correct_labels.shape, cm_correct_labels)\nprint(\"Predicted labels: \", cm_predictions.shape, cm_predictions)\n\ncmat = confusion_matrix(cm_correct_labels, cm_predictions, labels=range(len(CLASSES)))\nscore = f1_score(cm_correct_labels, cm_predictions, labels=range(len(CLASSES)), average='macro')\nprecision = precision_score(cm_correct_labels, cm_predictions, labels=range(len(CLASSES)), average='macro')\nrecall = recall_score(cm_correct_labels, cm_predictions, labels=range(len(CLASSES)), average='macro')\ncmat = (cmat.T / cmat.sum(axis=1)).T # normalized\ndisplay_confusion_matrix(cmat, score, precision, recall)\nprint('f1 score: {:.3f}, precision: {:.3f}, recall: {:.3f}'.format(score, precision, recall))","execution_count":null,"outputs":[]},{"metadata":{"_cell_guid":"81ca92ca-ec06-44c7-a73b-8d58de8579d1","_uuid":"ab2f47bf-0543-4f66-8e8f-09ea32fd3a1c","scrolled":false,"trusted":false},"cell_type":"code","source":"#model = enetb7_model\nmodel = enetb7_model2\n#model = xcept_model\n#model = dense201_model\n\nhistory = model.fit(get_training_dataset_mixup(), \n                    steps_per_epoch=STEPS_PER_EPOCH, epochs=EPOCHS,\n                    validation_data=get_validation_dataset(),\n                    callbacks=[early_stopping_cb, lr_callback])\n \ndisplay_training_curves(history.history['loss'], history.history['val_loss'], 'loss', 211)\ndisplay_training_curves(history.history['categorical_accuracy'], history.history['val_categorical_accuracy'], 'accuracy', 212)\n\n# Confusion matrix\ncmdataset = get_validation_dataset(ordered=True) # since we are splitting the dataset and iterating separately on images and labels, order matters.\nimages_ds = cmdataset.map(lambda image, label: image)\nlabels_ds = cmdataset.map(lambda image, label: label).unbatch()\ncm_correct_labels = next(iter(labels_ds.batch(NUM_VALIDATION_IMAGES))).numpy() # get everything as one batch\ncm_correct_labels = np.argmax(cm_correct_labels, axis=-1)\ncm_probabilities = model.predict(images_ds)\ncm_predictions = np.argmax(cm_probabilities, axis=-1)\nprint(\"Correct   labels: \", cm_correct_labels.shape, cm_correct_labels)\nprint(\"Predicted labels: \", cm_predictions.shape, cm_predictions)\n\ncmat = confusion_matrix(cm_correct_labels, cm_predictions, labels=range(len(CLASSES)))\nscore = f1_score(cm_correct_labels, cm_predictions, labels=range(len(CLASSES)), average='macro')\nprecision = precision_score(cm_correct_labels, cm_predictions, labels=range(len(CLASSES)), average='macro')\nrecall = recall_score(cm_correct_labels, cm_predictions, labels=range(len(CLASSES)), average='macro')\ncmat = (cmat.T / cmat.sum(axis=1)).T # normalized\ndisplay_confusion_matrix(cmat, score, precision, recall)\nprint('f1 score: {:.3f}, precision: {:.3f}, recall: {:.3f}'.format(score, precision, recall))","execution_count":null,"outputs":[]},{"metadata":{"_cell_guid":"81ca92ca-ec06-44c7-a73b-8d58de8579d1","_uuid":"ab2f47bf-0543-4f66-8e8f-09ea32fd3a1c","scrolled":false,"trusted":false},"cell_type":"code","source":"#model = enetb7_model\n#model = xcept_model\nmodel = dense201_model\n\nhistory = model.fit(get_training_dataset(), \n                    steps_per_epoch=STEPS_PER_EPOCH, epochs=EPOCHS,\n                    validation_data=get_validation_dataset(),\n                    callbacks=[early_stopping_cb, lr_callback])\n \ndisplay_training_curves(history.history['loss'], history.history['val_loss'], 'loss', 211)\ndisplay_training_curves(history.history['categorical_accuracy'], history.history['val_categorical_accuracy'], 'accuracy', 212)\n\n# Confusion matrix\ncmdataset = get_validation_dataset(ordered=True) # since we are splitting the dataset and iterating separately on images and labels, order matters.\nimages_ds = cmdataset.map(lambda image, label: image)\nlabels_ds = cmdataset.map(lambda image, label: label).unbatch()\ncm_correct_labels = next(iter(labels_ds.batch(NUM_VALIDATION_IMAGES))).numpy() # get everything as one batch\ncm_correct_labels = np.argmax(cm_correct_labels, axis=-1)\ncm_probabilities = model.predict(images_ds)\ncm_predictions = np.argmax(cm_probabilities, axis=-1)\nprint(\"Correct   labels: \", cm_correct_labels.shape, cm_correct_labels)\nprint(\"Predicted labels: \", cm_predictions.shape, cm_predictions)\n\ncmat = confusion_matrix(cm_correct_labels, cm_predictions, labels=range(len(CLASSES)))\nscore = f1_score(cm_correct_labels, cm_predictions, labels=range(len(CLASSES)), average='macro')\nprecision = precision_score(cm_correct_labels, cm_predictions, labels=range(len(CLASSES)), average='macro')\nrecall = recall_score(cm_correct_labels, cm_predictions, labels=range(len(CLASSES)), average='macro')\ncmat = (cmat.T / cmat.sum(axis=1)).T # normalized\ndisplay_confusion_matrix(cmat, score, precision, recall)\nprint('f1 score: {:.3f}, precision: {:.3f}, recall: {:.3f}'.format(score, precision, recall))","execution_count":null,"outputs":[]},{"metadata":{"_cell_guid":"81ca92ca-ec06-44c7-a73b-8d58de8579d1","_uuid":"ab2f47bf-0543-4f66-8e8f-09ea32fd3a1c","scrolled":false,"trusted":false},"cell_type":"code","source":"#model = enetb7_model\n#model = xcept_model\nmodel = dense201_model2\n\nhistory = model.fit(get_training_dataset_mixup(), \n                    steps_per_epoch=STEPS_PER_EPOCH, epochs=EPOCHS,\n                    validation_data=get_validation_dataset(),\n                    callbacks=[early_stopping_cb, lr_callback])\n \ndisplay_training_curves(history.history['loss'], history.history['val_loss'], 'loss', 211)\ndisplay_training_curves(history.history['categorical_accuracy'], history.history['val_categorical_accuracy'], 'accuracy', 212)\n\n# Confusion matrix\ncmdataset = get_validation_dataset(ordered=True) # since we are splitting the dataset and iterating separately on images and labels, order matters.\nimages_ds = cmdataset.map(lambda image, label: image)\nlabels_ds = cmdataset.map(lambda image, label: label).unbatch()\ncm_correct_labels = next(iter(labels_ds.batch(NUM_VALIDATION_IMAGES))).numpy() # get everything as one batch\ncm_correct_labels = np.argmax(cm_correct_labels, axis=-1)\ncm_probabilities = model.predict(images_ds)\ncm_predictions = np.argmax(cm_probabilities, axis=-1)\nprint(\"Correct   labels: \", cm_correct_labels.shape, cm_correct_labels)\nprint(\"Predicted labels: \", cm_predictions.shape, cm_predictions)\n\ncmat = confusion_matrix(cm_correct_labels, cm_predictions, labels=range(len(CLASSES)))\nscore = f1_score(cm_correct_labels, cm_predictions, labels=range(len(CLASSES)), average='macro')\nprecision = precision_score(cm_correct_labels, cm_predictions, labels=range(len(CLASSES)), average='macro')\nrecall = recall_score(cm_correct_labels, cm_predictions, labels=range(len(CLASSES)), average='macro')\ncmat = (cmat.T / cmat.sum(axis=1)).T # normalized\ndisplay_confusion_matrix(cmat, score, precision, recall)\nprint('f1 score: {:.3f}, precision: {:.3f}, recall: {:.3f}'.format(score, precision, recall))","execution_count":null,"outputs":[]},{"metadata":{"_cell_guid":"c69a9260-5d4e-424e-ad86-cb871902d4a6","_uuid":"07a1ab95-bcf1-469e-9a54-ca89a134fa6e","trusted":false},"cell_type":"code","source":"\n# Confusion matrix\ncmdataset = get_validation_dataset(ordered=True) # since we are splitting the dataset and iterating separately on images and labels, order matters.\nimages_ds = cmdataset.map(lambda image, label: image)\nlabels_ds = cmdataset.map(lambda image, label: label).unbatch()\ncm_correct_labels = next(iter(labels_ds.batch(NUM_VALIDATION_IMAGES))).numpy() # get everything as one batch\ncm_correct_labels = np.argmax(cm_correct_labels, axis=-1)\n\nenetb7_modelcm_probabilities = enetb7_model.predict(images_ds)\nenetb7_model2cm_probabilities = enetb7_model2.predict(images_ds)\ndense201_modelcm_probabilities = dense201_model.predict(images_ds)\ndense201_model2cm_probabilities = dense201_model2.predict(images_ds)\ncm_probabilities = enetb7_modelcm_probabilities + enetb7_model2cm_probabilities + dense201_modelcm_probabilities + dense201_model2cm_probabilities\n\n#cm_probabilities = model.predict(images_ds)\ncm_predictions = np.argmax(cm_probabilities, axis=-1)\nprint(\"Correct   labels: \", cm_correct_labels.shape, cm_correct_labels)\nprint(\"Predicted labels: \", cm_predictions.shape, cm_predictions)\n\ncmat = confusion_matrix(cm_correct_labels, cm_predictions, labels=range(len(CLASSES)))\nscore = f1_score(cm_correct_labels, cm_predictions, labels=range(len(CLASSES)), average='macro')\nprecision = precision_score(cm_correct_labels, cm_predictions, labels=range(len(CLASSES)), average='macro')\nrecall = recall_score(cm_correct_labels, cm_predictions, labels=range(len(CLASSES)), average='macro')\ncmat = (cmat.T / cmat.sum(axis=1)).T # normalized\ndisplay_confusion_matrix(cmat, score, precision, recall)\nprint('f1 score: {:.3f}, precision: {:.3f}, recall: {:.3f}'.format(score, precision, recall))\n\nprint('run time: ', datetime.now() - start_time)","execution_count":null,"outputs":[]},{"metadata":{"_cell_guid":"976a088c-8365-4e2d-8554-bff96991ec16","_uuid":"63ebab08-e1c6-4eb8-af19-b3e55f6b938f","scrolled":false,"trusted":false},"cell_type":"code","source":"# Predictions\ntest_ds = get_test_dataset(ordered=True) # since we are splitting the dataset and iterating separately on images and ids, order matters.\n\nprint('Computing predictions...')\ntest_images_ds = test_ds.map(lambda image, idnum: image)\n\n#probabilities = model.predict(test_images_ds)\nenetb7_modelcm_probabilities = enetb7_model.predict(test_images_ds)\nenetb7_model2cm_probabilities = enetb7_model2.predict(test_images_ds)\n#xcept_modelcm_probabilities = xcept_model.predict(test_images_ds)\ndense201_modelcm_probabilities = dense201_model.predict(test_images_ds)\ndense201_model2cm_probabilities = dense201_model2.predict(test_images_ds)\nprobabilities = enetb7_modelcm_probabilities + enetb7_model2cm_probabilities + dense201_modelcm_probabilities + dense201_model2cm_probabilities\n\npredictions = np.argmax(probabilities, axis=-1)\nprint(predictions)\n\nprint('Generating submission.csv file...')\ntest_ids_ds = test_ds.map(lambda image, idnum: idnum).unbatch()\ntest_ids = next(iter(test_ids_ds.batch(NUM_TEST_IMAGES))).numpy().astype('U') # all in one batch\nnp.savetxt('submission.csv', np.rec.fromarrays([test_ids, predictions]), fmt=['%s', '%d'], delimiter=',', header='id,label', comments='')\n#!head submission.csv\nwith open('submission.csv') as myfile:\n    head = [next(myfile) for x in range(10)]\nprint(head)","execution_count":null,"outputs":[]},{"metadata":{"_cell_guid":"8b55d4be-49d3-4834-99dc-1f9f50991071","_uuid":"de1be895-e7e2-41b6-b5ab-730875d6d728","scrolled":false,"trusted":false},"cell_type":"code","source":"","execution_count":null,"outputs":[]},{"metadata":{"_cell_guid":"d55caf20-96d9-47d9-b27d-45ac9824e3db","_uuid":"3de3afbe-6944-4a7e-a943-45228dc7cb46","trusted":false},"cell_type":"code","source":"","execution_count":null,"outputs":[]},{"metadata":{"_cell_guid":"dd1deabe-8c3a-4029-89bb-dea1e097bbb4","_uuid":"195aa691-f36d-4e2a-86ce-df46b5a2ded6","trusted":false},"cell_type":"code","source":"","execution_count":null,"outputs":[]}],"metadata":{"kernelspec":{"display_name":"Python 3","language":"python","name":"python3"},"language_info":{"codemirror_mode":{"name":"ipython","version":3},"file_extension":".py","mimetype":"text/x-python","name":"python","nbconvert_exporter":"python","pygments_lexer":"ipython3","version":"3.8.2"}},"nbformat":4,"nbformat_minor":4}