{"cells":[{"metadata":{"_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","execution":{"iopub.execute_input":"2020-11-03T21:29:02.913719Z","iopub.status.busy":"2020-11-03T21:29:02.912864Z","iopub.status.idle":"2020-11-03T21:29:09.845261Z","shell.execute_reply":"2020-11-03T21:29:09.846111Z"},"papermill":{"duration":6.965797,"end_time":"2020-11-03T21:29:09.8464","exception":false,"start_time":"2020-11-03T21:29:02.880603","status":"completed"},"tags":[],"trusted":true},"cell_type":"code","source":"import math, re, os\nimport tensorflow as tf\nimport numpy as np\nfrom matplotlib import pyplot as plt\nfrom kaggle_datasets import KaggleDatasets\nfrom sklearn.metrics import f1_score, precision_score, recall_score, confusion_matrix\nprint(\"Tensorflow version \" + tf.__version__)\nAUTO = tf.data.experimental.AUTOTUNE","execution_count":null,"outputs":[]},{"metadata":{"papermill":{"duration":0.022559,"end_time":"2020-11-03T21:29:09.893613","exception":false,"start_time":"2020-11-03T21:29:09.871054","status":"completed"},"tags":[]},"cell_type":"markdown","source":"# TPU or GPU detection"},{"metadata":{"execution":{"iopub.execute_input":"2020-11-03T21:29:10.047633Z","iopub.status.busy":"2020-11-03T21:29:09.955836Z","iopub.status.idle":"2020-11-03T21:29:14.204006Z","shell.execute_reply":"2020-11-03T21:29:14.203192Z"},"papermill":{"duration":4.287289,"end_time":"2020-11-03T21:29:14.204139","exception":false,"start_time":"2020-11-03T21:29:09.91685","status":"completed"},"tags":[],"trusted":true},"cell_type":"code","source":"try: # detect TPUs\n    tpu = tf.distribute.cluster_resolver.TPUClusterResolver() # TPU detection\n    tf.config.experimental_connect_to_cluster(tpu)\n    tf.tpu.experimental.initialize_tpu_system(tpu)\n    strategy = tf.distribute.experimental.TPUStrategy(tpu)\nexcept ValueError: # no TPU found, detect GPUs\n    strategy = tf.distribute.MirroredStrategy() # for GPU or multi-GPU machines\n    #strategy = tf.distribute.get_strategy() # default strategy that works on CPU and single GPU\n    #strategy = tf.distribute.experimental.MultiWorkerMirroredStrategy() # for clusters of multi-GPU machines\n\nprint(\"Number of accelerators: \", strategy.num_replicas_in_sync)","execution_count":null,"outputs":[]},{"metadata":{"papermill":{"duration":0.025298,"end_time":"2020-11-03T21:29:14.253553","exception":false,"start_time":"2020-11-03T21:29:14.228255","status":"completed"},"tags":[]},"cell_type":"markdown","source":"# Competition data access\nTPUs read data directly from Google Cloud Storage (GCS). This Kaggle utility will copy the dataset to a GCS bucket co-located with the TPU. If you have multiple datasets attached to the notebook, you can pass the name of a specific dataset to the get_gcs_path function. The name of the dataset is the name of the directory it is mounted in. Use `!ls /kaggle/input/` to list attached datasets."},{"metadata":{"execution":{"iopub.execute_input":"2020-11-03T21:29:14.320263Z","iopub.status.busy":"2020-11-03T21:29:14.319441Z","iopub.status.idle":"2020-11-03T21:29:14.638036Z","shell.execute_reply":"2020-11-03T21:29:14.637232Z"},"papermill":{"duration":0.360959,"end_time":"2020-11-03T21:29:14.638166","exception":false,"start_time":"2020-11-03T21:29:14.277207","status":"completed"},"tags":[],"trusted":true},"cell_type":"code","source":"GCS_DS_PATH = KaggleDatasets().get_gcs_path() # you can list the bucket with \"!gsutil ls $GCS_DS_PATH\"","execution_count":null,"outputs":[]},{"metadata":{"papermill":{"duration":0.023466,"end_time":"2020-11-03T21:29:14.685578","exception":false,"start_time":"2020-11-03T21:29:14.662112","status":"completed"},"tags":[]},"cell_type":"markdown","source":"# Configuration"},{"metadata":{"execution":{"iopub.execute_input":"2020-11-03T21:29:14.917292Z","iopub.status.busy":"2020-11-03T21:29:14.741879Z","iopub.status.idle":"2020-11-03T21:29:14.992087Z","shell.execute_reply":"2020-11-03T21:29:14.99118Z"},"papermill":{"duration":0.283024,"end_time":"2020-11-03T21:29:14.992228","exception":false,"start_time":"2020-11-03T21:29:14.709204","status":"completed"},"tags":[],"trusted":true},"cell_type":"code","source":"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.\nEPOCHS = 50\nBATCH_SIZE = 16 * 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":{"papermill":{"duration":0.024215,"end_time":"2020-11-03T21:29:15.040786","exception":false,"start_time":"2020-11-03T21:29:15.016571","status":"completed"},"tags":[]},"cell_type":"markdown","source":"## Visualization utilities\ndata -> pixels, nothing of much interest for the machine learning practitioner in this section."},{"metadata":{"execution":{"iopub.execute_input":"2020-11-03T21:29:15.133781Z","iopub.status.busy":"2020-11-03T21:29:15.103461Z","iopub.status.idle":"2020-11-03T21:29:15.137933Z","shell.execute_reply":"2020-11-03T21:29:15.137064Z"},"papermill":{"duration":0.073368,"end_time":"2020-11-03T21:29:15.138073","exception":false,"start_time":"2020-11-03T21:29:15.064705","status":"completed"},"tags":[],"trusted":true},"cell_type":"code","source":"# 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    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\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        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":{"papermill":{"duration":0.023696,"end_time":"2020-11-03T21:29:15.185958","exception":false,"start_time":"2020-11-03T21:29:15.162262","status":"completed"},"tags":[]},"cell_type":"markdown","source":"# Datasets"},{"metadata":{"_cell_guid":"79c7e3d0-c299-4dcb-8224-4455121ee9b0","_uuid":"d629ff2d2480ee46fbb7e2d37f6b5fab8052498a","execution":{"iopub.execute_input":"2020-11-03T21:29:15.266082Z","iopub.status.busy":"2020-11-03T21:29:15.265007Z","iopub.status.idle":"2020-11-03T21:29:15.268777Z","shell.execute_reply":"2020-11-03T21:29:15.269565Z"},"papermill":{"duration":0.05973,"end_time":"2020-11-03T21:29:15.269777","exception":false,"start_time":"2020-11-03T21:29:15.210047","status":"completed"},"tags":[],"trusted":true},"cell_type":"code","source":"def to_float32(image, label):\n    return tf.cast(image, tf.float32), label\n\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_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_flip_up_down(image)\n#     image = tf.image.rot90(image)\n\n    image = tf.image.rot90(image, tf.random.uniform(shape=[], minval=0, maxval=4, dtype=tf.dtypes.int32))\n    \n    minval = IMAGE_SIZE[0] // 2\n    maxval = IMAGE_SIZE[0]\n    random_shape = tf.random.uniform(shape=[2], minval=minval, maxval=maxval, dtype=tf.dtypes.int32)\n    image = tf.image.random_crop(image, size=[random_shape[0], random_shape[1], 3])\n    image = tf.image.resize(image, size=[IMAGE_SIZE[0], IMAGE_SIZE[1]])\n    \n#     if random_shape[0] < IMAGE_SIZE[0] or random_shape[1] < IMAGE_SIZE[0]:\n#         image = tf.image.resize(image, size=[IMAGE_SIZE[0], IMAGE_SIZE[1]])\n#     else:\n#         image = tf.image.random_crop(image, size=[IMAGE_SIZE[0], IMAGE_SIZE[1], 3])\n    \n    image = tf.image.random_saturation(image, 0.7, 1.3)\n    image = tf.image.random_brightness(image, 0.05)\n\n    image = tf.clip_by_value(image, 0, 1)\n    \n    return image, label   \n\ndef get_training_dataset():\n    dataset = load_dataset(TRAINING_FILENAMES, labeled=True)\n    dataset = dataset.map(data_augment, num_parallel_calls=AUTO)\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(ordered=False):\n    dataset = load_dataset(VALIDATION_FILENAMES, labeled=True, ordered=ordered)\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))","execution_count":null,"outputs":[]},{"metadata":{"papermill":{"duration":0.025058,"end_time":"2020-11-03T21:29:15.320917","exception":false,"start_time":"2020-11-03T21:29:15.295859","status":"completed"},"tags":[]},"cell_type":"markdown","source":"# Dataset visualizations"},{"metadata":{"execution":{"iopub.execute_input":"2020-11-03T21:29:15.385253Z","iopub.status.busy":"2020-11-03T21:29:15.383984Z","iopub.status.idle":"2020-11-03T21:29:23.143477Z","shell.execute_reply":"2020-11-03T21:29:23.142794Z"},"papermill":{"duration":7.797564,"end_time":"2020-11-03T21:29:23.143627","exception":false,"start_time":"2020-11-03T21:29:15.346063","status":"completed"},"tags":[],"trusted":true},"cell_type":"code","source":"# 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","execution_count":null,"outputs":[]},{"metadata":{"execution":{"iopub.execute_input":"2020-11-03T21:29:23.209576Z","iopub.status.busy":"2020-11-03T21:29:23.208454Z","iopub.status.idle":"2020-11-03T21:29:23.257565Z","shell.execute_reply":"2020-11-03T21:29:23.256838Z"},"papermill":{"duration":0.08484,"end_time":"2020-11-03T21:29:23.257697","exception":false,"start_time":"2020-11-03T21:29:23.172857","status":"completed"},"tags":[],"trusted":true},"cell_type":"code","source":"# Peek at training data\ntraining_dataset = get_training_dataset()\ntraining_dataset = training_dataset.unbatch().batch(20)\ntrain_batch = iter(training_dataset)","execution_count":null,"outputs":[]},{"metadata":{"execution":{"iopub.execute_input":"2020-11-03T21:29:23.65081Z","iopub.status.busy":"2020-11-03T21:29:23.321037Z","iopub.status.idle":"2020-11-03T21:29:25.756553Z","shell.execute_reply":"2020-11-03T21:29:25.757192Z"},"papermill":{"duration":2.47039,"end_time":"2020-11-03T21:29:25.757384","exception":false,"start_time":"2020-11-03T21:29:23.286994","status":"completed"},"tags":[],"trusted":true},"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":{"trusted":true},"cell_type":"code","source":"# Peek at validation data\nvalidation_dataset = get_validation_dataset()\nvalidation_dataset = validation_dataset.unbatch().batch(20)\nvalidation_batch = iter(validation_dataset)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"# run this cell again for next set of images\ndisplay_batch_of_images(next(validation_batch))","execution_count":null,"outputs":[]},{"metadata":{"execution":{"iopub.execute_input":"2020-11-03T21:29:25.894746Z","iopub.status.busy":"2020-11-03T21:29:25.89358Z","iopub.status.idle":"2020-11-03T21:29:25.927252Z","shell.execute_reply":"2020-11-03T21:29:25.926468Z"},"papermill":{"duration":0.10489,"end_time":"2020-11-03T21:29:25.927399","exception":false,"start_time":"2020-11-03T21:29:25.822509","status":"completed"},"tags":[],"trusted":true},"cell_type":"code","source":"# peer at test data\ntest_dataset = get_test_dataset()\ntest_dataset = test_dataset.unbatch().batch(20)\ntest_batch = iter(test_dataset)","execution_count":null,"outputs":[]},{"metadata":{"execution":{"iopub.execute_input":"2020-11-03T21:29:26.170445Z","iopub.status.busy":"2020-11-03T21:29:26.058509Z","iopub.status.idle":"2020-11-03T21:29:28.524856Z","shell.execute_reply":"2020-11-03T21:29:28.525472Z"},"papermill":{"duration":2.535039,"end_time":"2020-11-03T21:29:28.525631","exception":false,"start_time":"2020-11-03T21:29:25.990592","status":"completed"},"tags":[],"trusted":true},"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":{"papermill":{"duration":0.099992,"end_time":"2020-11-03T21:29:28.725918","exception":false,"start_time":"2020-11-03T21:29:28.625926","status":"completed"},"tags":[]},"cell_type":"markdown","source":"# Model"},{"metadata":{"execution":{"iopub.execute_input":"2020-11-03T21:29:28.93783Z","iopub.status.busy":"2020-11-03T21:29:28.936725Z","iopub.status.idle":"2020-11-03T21:29:32.552847Z","shell.execute_reply":"2020-11-03T21:29:32.552067Z"},"papermill":{"duration":3.726679,"end_time":"2020-11-03T21:29:32.553004","exception":false,"start_time":"2020-11-03T21:29:28.826325","status":"completed"},"tags":[],"trusted":true},"cell_type":"code","source":"from tensorflow.keras.models import Sequential\nfrom tensorflow.keras.layers import Conv2D, MaxPooling2D, GlobalAveragePooling2D, Dropout, Flatten, Dense\nfrom tensorflow.keras.models import Model\nfrom tensorflow.keras.applications import DenseNet201\n\nwith strategy.scope():\n    dense_net = DenseNet201(weights=\"imagenet\", include_top=False, input_shape=(IMAGE_SIZE[0], IMAGE_SIZE[1], 3))\n    \n#     for layer in dense_net.layers:\n#         layer.trainable = False\n    \n    top_model = dense_net.output\n    top_model = GlobalAveragePooling2D()(top_model)\n#     top_model = Flatten()(top_model)\n#     top_model = Dense(1024, activation=\"relu\")(top_model)\n#     top_model = Dense(1024, activation=\"relu\")(top_model)\n#     top_model = Dense(512, activation=\"relu\")(top_model)\n    top_model = Dense(len(CLASSES), activation=\"softmax\")(top_model)\n    \n    model = Model(inputs=dense_net.input, outputs = top_model)\n\n    model.compile(optimizer=tf.keras.optimizers.Adam(), loss = 'sparse_categorical_crossentropy', metrics=['accuracy'])\nmodel.summary()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"!pip install visualkeras\nimport visualkeras\n\nvisualkeras.layered_view(model)","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"# Custom LR scheduler"},{"metadata":{"trusted":true},"cell_type":"code","source":"# Using an LR ramp up because fine-tuning a pre-trained model.\n# Starting with a high LR would break the pre-trained weights.\n\nLR_START = 0.00001\nLR_MAX = 0.00005 * strategy.num_replicas_in_sync\nLR_MIN = 0.00001\nLR_RAMPUP_EPOCHS = 10\nLR_SUSTAIN_EPOCHS = 0\nLR_EXP_DECAY = 0.9\n\ndef lrfn(epoch):\n    if epoch < LR_RAMPUP_EPOCHS:\n        lr = (LR_MAX - LR_START) / LR_RAMPUP_EPOCHS * epoch + LR_START\n    elif epoch < LR_RAMPUP_EPOCHS + LR_SUSTAIN_EPOCHS:\n        lr = LR_MAX\n    else:\n        lr = (LR_MAX - LR_MIN) * LR_EXP_DECAY**(epoch - LR_RAMPUP_EPOCHS - LR_SUSTAIN_EPOCHS) + LR_MIN\n    return lr\n\nrng = [i for i in range(25 if EPOCHS<25 else EPOCHS)]\ny = [lrfn(x) for x in rng]\nplt.plot(rng, y)\nprint(\"Learning rate schedule: {:.3g} to {:.3g} to {:.3g}\".format(y[0], max(y), y[-1]))","execution_count":null,"outputs":[]},{"metadata":{"papermill":{"duration":0.103908,"end_time":"2020-11-03T21:29:32.763886","exception":false,"start_time":"2020-11-03T21:29:32.659978","status":"completed"},"tags":[]},"cell_type":"markdown","source":"# Training"},{"metadata":{"execution":{"iopub.execute_input":"2020-11-03T21:29:32.987072Z","iopub.status.busy":"2020-11-03T21:29:32.985954Z","iopub.status.idle":"2020-11-03T21:34:07.055989Z","shell.execute_reply":"2020-11-03T21:34:07.055136Z"},"papermill":{"duration":274.184006,"end_time":"2020-11-03T21:34:07.056142","exception":false,"start_time":"2020-11-03T21:29:32.872136","status":"completed"},"tags":[],"trusted":true},"cell_type":"code","source":"from tensorflow.keras.callbacks import ModelCheckpoint, EarlyStopping, ReduceLROnPlateau, LearningRateScheduler\n\n# TPUs need images in float format\ntraining_dataset = get_training_dataset().map(to_float32)\nvalidation_dataset = get_validation_dataset().map(to_float32)\n\ncheckpoint = ModelCheckpoint(\"model_checkpoint.h5\", monitor='val_loss', mode='min', save_best_only=True, verbose=1)\nearly_stopping = EarlyStopping(monitor='val_loss', min_delta=0, patience=10, verbose=1, restore_best_weights=True)\nreduce_lr = ReduceLROnPlateau(monitor='val_loss', factor=0.2, patience=3, verbose=1, min_delta=0.0001)\nschedule_lr = LearningRateScheduler(lrfn, verbose=1)\n\ncallbacks = [early_stopping, schedule_lr]\n\nhistory = model.fit(training_dataset, steps_per_epoch=STEPS_PER_EPOCH, epochs=EPOCHS, validation_data=validation_dataset, callbacks=callbacks)","execution_count":null,"outputs":[]},{"metadata":{"execution":{"iopub.execute_input":"2020-11-03T21:34:08.249239Z","iopub.status.busy":"2020-11-03T21:34:08.24846Z","iopub.status.idle":"2020-11-03T21:34:08.731509Z","shell.execute_reply":"2020-11-03T21:34:08.732208Z"},"papermill":{"duration":1.07837,"end_time":"2020-11-03T21:34:08.732405","exception":false,"start_time":"2020-11-03T21:34:07.654035","status":"completed"},"tags":[],"trusted":true},"cell_type":"code","source":"display_training_curves(history.history['loss'], history.history['val_loss'], 'loss', 211)\ndisplay_training_curves(history.history['accuracy'], history.history['val_accuracy'], 'accuracy', 212)","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"# Save the model"},{"metadata":{"trusted":true},"cell_type":"code","source":"# Save the model\nmodel.save('model.h5')","execution_count":null,"outputs":[]},{"metadata":{"papermill":{"duration":0.607641,"end_time":"2020-11-03T21:34:09.937509","exception":false,"start_time":"2020-11-03T21:34:09.329868","status":"completed"},"tags":[]},"cell_type":"markdown","source":"# Confusion matrix"},{"metadata":{"execution":{"iopub.execute_input":"2020-11-03T21:34:11.18653Z","iopub.status.busy":"2020-11-03T21:34:11.18548Z","iopub.status.idle":"2020-11-03T21:34:26.240095Z","shell.execute_reply":"2020-11-03T21:34:26.239266Z"},"papermill":{"duration":15.712351,"end_time":"2020-11-03T21:34:26.24024","exception":false,"start_time":"2020-11-03T21:34:10.527889","status":"completed"},"tags":[],"trusted":true},"cell_type":"code","source":"cmdataset = get_validation_dataset(ordered=True) # since we are splitting the dataset and iterating separately on images and labels, order matters.\ncmdataset = cmdataset.map(to_float32)\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_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)","execution_count":null,"outputs":[]},{"metadata":{"execution":{"iopub.execute_input":"2020-11-03T21:34:27.437031Z","iopub.status.busy":"2020-11-03T21:34:27.436232Z","iopub.status.idle":"2020-11-03T21:34:29.826448Z","shell.execute_reply":"2020-11-03T21:34:29.825764Z"},"papermill":{"duration":2.992853,"end_time":"2020-11-03T21:34:29.826581","exception":false,"start_time":"2020-11-03T21:34:26.833728","status":"completed"},"tags":[],"trusted":true},"cell_type":"code","source":"cmat = 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":{"papermill":{"duration":0.75004,"end_time":"2020-11-03T21:34:31.177024","exception":false,"start_time":"2020-11-03T21:34:30.426984","status":"completed"},"tags":[]},"cell_type":"markdown","source":"# Predictions"},{"metadata":{"execution":{"iopub.execute_input":"2020-11-03T21:34:32.512267Z","iopub.status.busy":"2020-11-03T21:34:32.511482Z","iopub.status.idle":"2020-11-03T21:34:53.282007Z","shell.execute_reply":"2020-11-03T21:34:53.281118Z"},"papermill":{"duration":21.424172,"end_time":"2020-11-03T21:34:53.282156","exception":false,"start_time":"2020-11-03T21:34:31.857984","status":"completed"},"tags":[],"trusted":true},"cell_type":"code","source":"test_ds = get_test_dataset(ordered=True) # since we are splitting the dataset and iterating separately on images and ids, order matters.\ntest_ds = test_ds.map(to_float32)\n\nprint('Computing predictions...')\ntest_images_ds = test_ds.map(lambda image, idnum: image)\nprobabilities = model.predict(test_images_ds)\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","execution_count":null,"outputs":[]},{"metadata":{"papermill":{"duration":0.597497,"end_time":"2020-11-03T21:34:54.477817","exception":false,"start_time":"2020-11-03T21:34:53.88032","status":"completed"},"tags":[]},"cell_type":"markdown","source":"# Visual validation"},{"metadata":{"execution":{"iopub.execute_input":"2020-11-03T21:34:55.684157Z","iopub.status.busy":"2020-11-03T21:34:55.683328Z","iopub.status.idle":"2020-11-03T21:34:55.723868Z","shell.execute_reply":"2020-11-03T21:34:55.723209Z"},"papermill":{"duration":0.64953,"end_time":"2020-11-03T21:34:55.724013","exception":false,"start_time":"2020-11-03T21:34:55.074483","status":"completed"},"tags":[],"trusted":true},"cell_type":"code","source":"dataset = get_validation_dataset()\ndataset = dataset.unbatch().batch(20)\nbatch = iter(dataset)","execution_count":null,"outputs":[]},{"metadata":{"execution":{"iopub.execute_input":"2020-11-03T21:34:56.939228Z","iopub.status.busy":"2020-11-03T21:34:56.938071Z","iopub.status.idle":"2020-11-03T21:35:14.946671Z","shell.execute_reply":"2020-11-03T21:35:14.947317Z"},"papermill":{"duration":18.620488,"end_time":"2020-11-03T21:35:14.947505","exception":false,"start_time":"2020-11-03T21:34:56.327017","status":"completed"},"tags":[],"trusted":true},"cell_type":"code","source":"# run this cell again for next set of images\nimages, labels = next(batch)\nprobabilities = model.predict(tf.cast(images, tf.float32))\npredictions = np.argmax(probabilities, axis=-1)\ndisplay_batch_of_images((images, labels), predictions)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"","execution_count":null,"outputs":[]}],"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":4,"nbformat_minor":4}