{"cells":[{"metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","trusted":true},"cell_type":"code","source":"batch_size = 4096\nSTROKE_COUNT = 196\nTRAIN_SAMPLES = 750\nVALID_SAMPLES = 75\nTEST_SAMPLES = 50","execution_count":19,"outputs":[]},{"metadata":{"_cell_guid":"79c7e3d0-c299-4dcb-8224-4455121ee9b0","_uuid":"d629ff2d2480ee46fbb7e2d37f6b5fab8052498a","trusted":true},"cell_type":"code","source":"%matplotlib inline\nimport os\nimport numpy as np\nimport matplotlib.pyplot as plt\nfrom keras.utils.np_utils import to_categorical\nfrom keras.preprocessing.sequence import pad_sequences\nfrom sklearn.preprocessing import LabelEncoder\nimport pandas as pd\nfrom keras.metrics import top_k_categorical_accuracy\ndef top_3_accuracy(x,y): return top_k_categorical_accuracy(x,y, 3)\nfrom keras.callbacks import ModelCheckpoint, LearningRateScheduler, EarlyStopping, ReduceLROnPlateau\nfrom glob import glob\nimport gc\ngc.enable()\ndef get_available_gpus():\n    from tensorflow.python.client import device_lib\n    local_device_protos = device_lib.list_local_devices()\n    return [x.name for x in local_device_protos if x.device_type == 'GPU']\nbase_dir = os.path.join('..', 'input')\ntest_path = os.path.join(base_dir, 'test_simplified.csv')","execution_count":20,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"from ast import literal_eval\nALL_TRAIN_PATHS = glob(os.path.join(base_dir, 'train_simplified', '*.csv'))\nCOL_NAMES = ['countrycode', 'drawing', 'key_id', 'recognized', 'timestamp', 'word']\n\ndef _stack_it(raw_strokes):\n    \"\"\"preprocess the string and make \n    a standard Nx3 stroke vector\"\"\"\n    stroke_vec = literal_eval(raw_strokes) # string->list\n    # unwrap the list\n    in_strokes = [(xi,yi,i)  \n     for i,(x,y) in enumerate(stroke_vec) \n     for xi,yi in zip(x,y)]\n    c_strokes = np.stack(in_strokes)\n    # replace stroke id with 1 for continue, 2 for new\n    c_strokes[:,2] = [1]+np.diff(c_strokes[:,2]).tolist()\n    c_strokes[:,2] += 1 # since 0 is no stroke\n    # pad the strokes with zeros\n    return pad_sequences(c_strokes.swapaxes(0, 1), \n                         maxlen=STROKE_COUNT, \n                         padding='post').swapaxes(0, 1)\ndef read_batch(samples=5, \n               start_row=0,\n               max_rows = 1000):\n    \"\"\"\n    load and process the csv files\n    this function is horribly inefficient but simple\n    \"\"\"\n    out_df_list = []\n    for c_path in ALL_TRAIN_PATHS:\n        c_df = pd.read_csv(c_path, nrows=max_rows, skiprows=start_row)\n        c_df.columns=COL_NAMES\n        out_df_list += [c_df.sample(samples)[['drawing', 'word']]]\n    full_df = pd.concat(out_df_list)\n    full_df['drawing'] = full_df['drawing'].\\\n        map(_stack_it)\n    \n    return full_df","execution_count":21,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"train_args = dict(samples=TRAIN_SAMPLES, \n                  start_row=0, \n                  max_rows=int(TRAIN_SAMPLES*1.5))\nvalid_args = dict(samples=VALID_SAMPLES, \n                  start_row=train_args['max_rows']+1, \n                  max_rows=VALID_SAMPLES+25)\ntest_args = dict(samples=TEST_SAMPLES, \n                 start_row=valid_args['max_rows']+train_args['max_rows']+1, \n                 max_rows=TEST_SAMPLES+25)\ntrain_df = read_batch(**train_args)\nvalid_df = read_batch(**valid_args)\ntest_df = read_batch(**test_args)\nword_encoder = LabelEncoder()\nword_encoder.fit(train_df['word'])\nprint('words', len(word_encoder.classes_), '=>', ', '.join([x for x in word_encoder.classes_]))","execution_count":22,"outputs":[{"output_type":"stream","text":"words 340 => The Eiffel Tower, The Great Wall of China, The Mona Lisa, airplane, alarm clock, ambulance, angel, animal migration, ant, anvil, apple, arm, asparagus, axe, backpack, banana, bandage, barn, baseball, baseball bat, basket, basketball, bat, bathtub, beach, bear, beard, bed, bee, belt, bench, bicycle, binoculars, bird, birthday cake, blackberry, blueberry, book, boomerang, bottlecap, bowtie, bracelet, brain, bread, bridge, broccoli, broom, bucket, bulldozer, bus, bush, butterfly, cactus, cake, calculator, calendar, camel, camera, camouflage, campfire, candle, cannon, canoe, car, carrot, castle, cat, ceiling fan, cell phone, cello, chair, chandelier, church, circle, clarinet, clock, cloud, coffee cup, compass, computer, cookie, cooler, couch, cow, crab, crayon, crocodile, crown, cruise ship, cup, diamond, dishwasher, diving board, dog, dolphin, donut, door, dragon, dresser, drill, drums, duck, dumbbell, ear, elbow, elephant, envelope, eraser, eye, eyeglasses, face, fan, feather, fence, finger, fire hydrant, fireplace, firetruck, fish, flamingo, flashlight, flip flops, floor lamp, flower, flying saucer, foot, fork, frog, frying pan, garden, garden hose, giraffe, goatee, golf club, grapes, grass, guitar, hamburger, hammer, hand, harp, hat, headphones, hedgehog, helicopter, helmet, hexagon, hockey puck, hockey stick, horse, hospital, hot air balloon, hot dog, hot tub, hourglass, house, house plant, hurricane, ice cream, jacket, jail, kangaroo, key, keyboard, knee, ladder, lantern, laptop, leaf, leg, light bulb, lighthouse, lightning, line, lion, lipstick, lobster, lollipop, mailbox, map, marker, matches, megaphone, mermaid, microphone, microwave, monkey, moon, mosquito, motorbike, mountain, mouse, moustache, mouth, mug, mushroom, nail, necklace, nose, ocean, octagon, octopus, onion, oven, owl, paint can, paintbrush, palm tree, panda, pants, paper clip, parachute, parrot, passport, peanut, pear, peas, pencil, penguin, piano, pickup truck, picture frame, pig, pillow, pineapple, pizza, pliers, police car, pond, pool, popsicle, postcard, potato, power outlet, purse, rabbit, raccoon, radio, rain, rainbow, rake, remote control, rhinoceros, river, roller coaster, rollerskates, sailboat, sandwich, saw, saxophone, school bus, scissors, scorpion, screwdriver, sea turtle, see saw, shark, sheep, shoe, shorts, shovel, sink, skateboard, skull, skyscraper, sleeping bag, smiley face, snail, snake, snorkel, snowflake, snowman, soccer ball, sock, speedboat, spider, spoon, spreadsheet, square, squiggle, squirrel, stairs, star, steak, stereo, stethoscope, stitches, stop sign, stove, strawberry, streetlight, string bean, submarine, suitcase, sun, swan, sweater, swing set, sword, t-shirt, table, teapot, teddy-bear, telephone, television, tennis racquet, tent, tiger, toaster, toe, toilet, tooth, toothbrush, toothpaste, tornado, tractor, traffic light, train, tree, triangle, trombone, truck, trumpet, umbrella, underwear, van, vase, violin, washing machine, watermelon, waterslide, whale, wheel, windmill, wine bottle, wine glass, wristwatch, yoga, zebra, zigzag\n","name":"stdout"}]},{"metadata":{"trusted":true},"cell_type":"code","source":"","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"def get_Xy(in_df):\n    #print(\"in_df shape :\", in_df.shape) # (255000,2) for train&vaild, (17000,2) for test\n    #print(in_df.iloc[0][0].shape) #(196,3) \n    X = np.stack(in_df['drawing'], 0)\n    #print(\"X shape : \", X.shape) #(255000,196,3) for train&vaild, (17000,196,3) for test\n    y = to_categorical(word_encoder.transform(in_df['word'].values))\n    #print(y.shape) # (255000, 340) for train&vaild, (17000,340) for test \n    # 340 = len(word_encoder.classes_)\n    return X, y\ntrain_X, train_y = get_Xy(train_df)\nvalid_X, valid_y = get_Xy(valid_df)\ntest_X, test_y = get_Xy(test_df)\nprint(train_X.shape)","execution_count":44,"outputs":[{"output_type":"stream","text":"(255000, 196, 3)\n","name":"stdout"}]},{"metadata":{"trusted":true},"cell_type":"code","source":"ig, m_axs = plt.subplots(3,3, figsize = (16, 16))\nprint(\"ig\", ig) # Figure(1152x1152)\nprint(\"m_axs.shape\", m_axs.shape) # (3, 3)\nrand_idxs = np.random.choice(range(train_X.shape[0]), size = 9) \nprint(\"rand_idxs.shape \", rand_idxs.shape) # (9,)\nfor c_id, c_ax in zip(rand_idxs, m_axs.flatten()):\n    test_arr = train_X[c_id]\n    print(\"train_X[c_id].shape\",train_X[c_id].shape) # (196, 3)\n    test_arr = test_arr[test_arr[:,2]>0, :] # only keep valid points\n    #print(\"test_arr : \", test_arr)\n    print(\"test_arr.shape : \", test_arr.shape) # (?,3)\n    lab_idx = np.cumsum(test_arr[:,2]-1)\n    #print(\"lab_idx : \", lab_idx) \n    print(\"lab_idx.shape : \", lab_idx.shape) # (/,)\n    for i in np.unique(lab_idx):\n        c_ax.plot(test_arr[lab_idx==i,0], \n                np.max(test_arr[:,1])-test_arr[lab_idx==i,1], '.-') \n    c_ax.axis('off')\n    c_ax.set_title(word_encoder.classes_[np.argmax(train_y[c_id])])","execution_count":62,"outputs":[{"output_type":"stream","text":"ig Figure(1152x1152)\nm_axs.shape (3, 3)\nrand_idxs.shape  (9,)\ntrain_X[c_id].shape (196, 3)\ntest_arr.shape :  (18, 3)\nlab_idx.shape :  (18,)\ntrain_X[c_id].shape (196, 3)\ntest_arr.shape :  (25, 3)\nlab_idx.shape :  (25,)\ntrain_X[c_id].shape (196, 3)\ntest_arr.shape :  (37, 3)\nlab_idx.shape :  (37,)\ntrain_X[c_id].shape (196, 3)\ntest_arr.shape :  (32, 3)\nlab_idx.shape :  (32,)\ntrain_X[c_id].shape (196, 3)\ntest_arr.shape :  (31, 3)\nlab_idx.shape :  (31,)\ntrain_X[c_id].shape (196, 3)\ntest_arr.shape :  (61, 3)\nlab_idx.shape :  (61,)\ntrain_X[c_id].shape (196, 3)\ntest_arr.shape :  (35, 3)\nlab_idx.shape :  (35,)\ntrain_X[c_id].shape (196, 3)\ntest_arr.shape :  (40, 3)\nlab_idx.shape :  (40,)\ntrain_X[c_id].shape (196, 3)\ntest_arr.shape :  (37, 3)\nlab_idx.shape :  (37,)\n","name":"stdout"},{"output_type":"display_data","data":{"text/plain":"<Figure size 1152x1152 with 9 Axes>","image/png":"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\n"},"metadata":{}}]},{"metadata":{"trusted":true},"cell_type":"code","source":"from keras.models import Sequential\nfrom keras.layers import BatchNormalization, Conv1D, LSTM, Dense, Dropout\nif len(get_available_gpus())>0:\n    # https://twitter.com/fchollet/status/918170264608817152?lang=en\n    from keras.layers import CuDNNLSTM as LSTM # this one is about 3x faster on GPU instances\nstroke_read_model = Sequential()\nstroke_read_model.add(BatchNormalization(input_shape = (None,)+train_X.shape[2:]))\n# train_X.shape[2:] = (254998, 196, 3)\n# (None,) + train_X.shape[2:] = (None, 254998, 196, 3)\n# filter count and length are taken from the script https://github.com/tensorflow/models/blob/master/tutorials/rnn/quickdraw/train_model.py\nstroke_read_model.add(Conv1D(48, (5,)))\nstroke_read_model.add(Dropout(0.3))\nstroke_read_model.add(Conv1D(64, (5,)))\nstroke_read_model.add(Dropout(0.3))\nstroke_read_model.add(Conv1D(96, (3,)))\nstroke_read_model.add(Dropout(0.3))\nstroke_read_model.add(LSTM(128, return_sequences = True))\nstroke_read_model.add(Dropout(0.3))\nstroke_read_model.add(LSTM(128, return_sequences = False))\nstroke_read_model.add(Dropout(0.3))\nstroke_read_model.add(Dense(512))\nstroke_read_model.add(Dropout(0.3))\nstroke_read_model.add(Dense(len(word_encoder.classes_), activation = 'softmax'))\nstroke_read_model.compile(optimizer = 'adam', \n                          loss = 'categorical_crossentropy', \n                          metrics = ['categorical_accuracy', top_3_accuracy])\nstroke_read_model.summary()","execution_count":63,"outputs":[{"output_type":"stream","text":"WARNING:tensorflow:From /opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/op_def_library.py:263: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.\nInstructions for updating:\nColocations handled automatically by placer.\nWARNING:tensorflow:From /opt/conda/lib/python3.6/site-packages/keras/backend/tensorflow_backend.py:3445: calling dropout (from tensorflow.python.ops.nn_ops) with keep_prob is deprecated and will be removed in a future version.\nInstructions for updating:\nPlease use `rate` instead of `keep_prob`. Rate should be set to `rate = 1 - keep_prob`.\n_________________________________________________________________\nLayer (type)                 Output Shape              Param #   \n=================================================================\nbatch_normalization_1 (Batch (None, None, 3)           12        \n_________________________________________________________________\nconv1d_1 (Conv1D)            (None, None, 48)          768       \n_________________________________________________________________\ndropout_1 (Dropout)          (None, None, 48)          0         \n_________________________________________________________________\nconv1d_2 (Conv1D)            (None, None, 64)          15424     \n_________________________________________________________________\ndropout_2 (Dropout)          (None, None, 64)          0         \n_________________________________________________________________\nconv1d_3 (Conv1D)            (None, None, 96)          18528     \n_________________________________________________________________\ndropout_3 (Dropout)          (None, None, 96)          0         \n_________________________________________________________________\ncu_dnnlstm_1 (CuDNNLSTM)     (None, None, 128)         115712    \n_________________________________________________________________\ndropout_4 (Dropout)          (None, None, 128)         0         \n_________________________________________________________________\ncu_dnnlstm_2 (CuDNNLSTM)     (None, 128)               132096    \n_________________________________________________________________\ndropout_5 (Dropout)          (None, 128)               0         \n_________________________________________________________________\ndense_1 (Dense)              (None, 512)               66048     \n_________________________________________________________________\ndropout_6 (Dropout)          (None, 512)               0         \n_________________________________________________________________\ndense_2 (Dense)              (None, 340)               174420    \n=================================================================\nTotal params: 523,008\nTrainable params: 523,002\nNon-trainable params: 6\n_________________________________________________________________\n","name":"stdout"}]},{"metadata":{"trusted":true},"cell_type":"code","source":"weight_path=\"{}_weights.best.hdf5\".format('stroke_lstm_model')\nprint(weight_path)\ncheckpoint = ModelCheckpoint(weight_path, monitor='val_loss', verbose=1, \n                             save_best_only=True, mode='min', save_weights_only = True)\n\n\nreduceLROnPlat = ReduceLROnPlateau(monitor='val_loss', factor=0.8, patience=10, \n                                   verbose=1, mode='auto', epsilon=0.0001, cooldown=5, min_lr=0.0001)\nearly = EarlyStopping(monitor=\"val_loss\", \n                      mode=\"min\", \n                      patience=5) # probably needs to be more patient, but kaggle time is limited\ncallbacks_list = [checkpoint, early, reduceLROnPlat]\n","execution_count":69,"outputs":[{"output_type":"stream","text":"stroke_lstm_model_weights.best.hdf5\n","name":"stdout"},{"output_type":"stream","text":"/opt/conda/lib/python3.6/site-packages/keras/callbacks.py:1065: UserWarning: `epsilon` argument is deprecated and will be removed, use `min_delta` instead.\n  warnings.warn('`epsilon` argument is deprecated and '\n","name":"stderr"}]},{"metadata":{"trusted":true},"cell_type":"code","source":"from IPython.display import clear_output\nstroke_read_model.fit(train_X, train_y,\n                      validation_data = (valid_X, valid_y), \n                      batch_size = batch_size,\n                      epochs = 50,\n                      callbacks = callbacks_list)\nstroke_read_model.summary()\n#clear_output()","execution_count":71,"outputs":[{"output_type":"stream","text":"Train on 255000 samples, validate on 25500 samples\nEpoch 1/50\n255000/255000 [==============================] - 24s 93us/step - loss: 2.4304 - categorical_accuracy: 0.4201 - top_3_accuracy: 0.6369 - val_loss: 2.1394 - val_categorical_accuracy: 0.4863 - val_top_3_accuracy: 0.6931\n\nEpoch 00001: val_loss improved from 2.17737 to 2.13941, saving model to stroke_lstm_model_weights.best.hdf5\nEpoch 2/50\n255000/255000 [==============================] - 24s 93us/step - loss: 2.4127 - categorical_accuracy: 0.4234 - top_3_accuracy: 0.6404 - val_loss: 2.1185 - val_categorical_accuracy: 0.4868 - val_top_3_accuracy: 0.6974\n\nEpoch 00002: val_loss improved from 2.13941 to 2.11852, saving model to stroke_lstm_model_weights.best.hdf5\nEpoch 3/50\n255000/255000 [==============================] - 24s 93us/step - loss: 2.3913 - categorical_accuracy: 0.4282 - top_3_accuracy: 0.6446 - val_loss: 2.1106 - val_categorical_accuracy: 0.4909 - val_top_3_accuracy: 0.7009\n\nEpoch 00003: val_loss improved from 2.11852 to 2.11059, saving model to stroke_lstm_model_weights.best.hdf5\nEpoch 4/50\n255000/255000 [==============================] - 24s 93us/step - loss: 2.3720 - categorical_accuracy: 0.4317 - top_3_accuracy: 0.6488 - val_loss: 2.1036 - val_categorical_accuracy: 0.4913 - val_top_3_accuracy: 0.7001\n\nEpoch 00004: val_loss improved from 2.11059 to 2.10361, saving model to stroke_lstm_model_weights.best.hdf5\nEpoch 5/50\n255000/255000 [==============================] - 24s 94us/step - loss: 2.3633 - categorical_accuracy: 0.4339 - top_3_accuracy: 0.6500 - val_loss: 2.0764 - val_categorical_accuracy: 0.4973 - val_top_3_accuracy: 0.7066\n\nEpoch 00005: val_loss improved from 2.10361 to 2.07638, saving model to stroke_lstm_model_weights.best.hdf5\nEpoch 6/50\n255000/255000 [==============================] - 24s 93us/step - loss: 2.3362 - categorical_accuracy: 0.4399 - top_3_accuracy: 0.6562 - val_loss: 2.0499 - val_categorical_accuracy: 0.5012 - val_top_3_accuracy: 0.7127\n\nEpoch 00006: val_loss improved from 2.07638 to 2.04990, saving model to stroke_lstm_model_weights.best.hdf5\nEpoch 7/50\n255000/255000 [==============================] - 24s 94us/step - loss: 2.3273 - categorical_accuracy: 0.4419 - top_3_accuracy: 0.6576 - val_loss: 2.0423 - val_categorical_accuracy: 0.5053 - val_top_3_accuracy: 0.7107\n\nEpoch 00007: val_loss improved from 2.04990 to 2.04232, saving model to stroke_lstm_model_weights.best.hdf5\nEpoch 8/50\n255000/255000 [==============================] - 24s 94us/step - loss: 2.3095 - categorical_accuracy: 0.4456 - top_3_accuracy: 0.6623 - val_loss: 2.0395 - val_categorical_accuracy: 0.5054 - val_top_3_accuracy: 0.7123\n\nEpoch 00008: val_loss improved from 2.04232 to 2.03954, saving model to stroke_lstm_model_weights.best.hdf5\nEpoch 9/50\n255000/255000 [==============================] - 24s 93us/step - loss: 2.2857 - categorical_accuracy: 0.4505 - top_3_accuracy: 0.6669 - val_loss: 2.0181 - val_categorical_accuracy: 0.5083 - val_top_3_accuracy: 0.7172\n\nEpoch 00009: val_loss improved from 2.03954 to 2.01811, saving model to stroke_lstm_model_weights.best.hdf5\nEpoch 10/50\n255000/255000 [==============================] - 24s 94us/step - loss: 2.2743 - categorical_accuracy: 0.4533 - top_3_accuracy: 0.6692 - val_loss: 2.0378 - val_categorical_accuracy: 0.5055 - val_top_3_accuracy: 0.7112\n\nEpoch 00010: val_loss did not improve from 2.01811\nEpoch 11/50\n255000/255000 [==============================] - 24s 94us/step - loss: 2.2649 - categorical_accuracy: 0.4544 - top_3_accuracy: 0.6706 - val_loss: 2.0197 - val_categorical_accuracy: 0.5075 - val_top_3_accuracy: 0.7153\n\nEpoch 00011: val_loss did not improve from 2.01811\nEpoch 12/50\n255000/255000 [==============================] - 24s 94us/step - loss: 2.2479 - categorical_accuracy: 0.4583 - top_3_accuracy: 0.6741 - val_loss: 1.9789 - val_categorical_accuracy: 0.5207 - val_top_3_accuracy: 0.7250\n\nEpoch 00012: val_loss improved from 2.01811 to 1.97890, saving model to stroke_lstm_model_weights.best.hdf5\nEpoch 13/50\n255000/255000 [==============================] - 24s 93us/step - loss: 2.2343 - categorical_accuracy: 0.4620 - top_3_accuracy: 0.6767 - val_loss: 1.9876 - val_categorical_accuracy: 0.5142 - val_top_3_accuracy: 0.7237\n\nEpoch 00013: val_loss did not improve from 1.97890\nEpoch 14/50\n255000/255000 [==============================] - 24s 93us/step - loss: 2.2211 - categorical_accuracy: 0.4633 - top_3_accuracy: 0.6793 - val_loss: 1.9843 - val_categorical_accuracy: 0.5162 - val_top_3_accuracy: 0.7244\n\nEpoch 00014: val_loss did not improve from 1.97890\nEpoch 15/50\n255000/255000 [==============================] - 24s 94us/step - loss: 2.2084 - categorical_accuracy: 0.4671 - top_3_accuracy: 0.6806 - val_loss: 1.9393 - val_categorical_accuracy: 0.5260 - val_top_3_accuracy: 0.7318\n\nEpoch 00015: val_loss improved from 1.97890 to 1.93929, saving model to stroke_lstm_model_weights.best.hdf5\nEpoch 16/50\n255000/255000 [==============================] - 24s 94us/step - loss: 2.1925 - categorical_accuracy: 0.4693 - top_3_accuracy: 0.6851 - val_loss: 1.9667 - val_categorical_accuracy: 0.5196 - val_top_3_accuracy: 0.7269\n\nEpoch 00016: val_loss did not improve from 1.93929\nEpoch 17/50\n255000/255000 [==============================] - 24s 94us/step - loss: 2.1818 - categorical_accuracy: 0.4719 - top_3_accuracy: 0.6872 - val_loss: 1.9761 - val_categorical_accuracy: 0.5167 - val_top_3_accuracy: 0.7245\n\nEpoch 00017: val_loss did not improve from 1.93929\nEpoch 18/50\n255000/255000 [==============================] - 24s 94us/step - loss: 2.1754 - categorical_accuracy: 0.4731 - top_3_accuracy: 0.6878 - val_loss: 1.9312 - val_categorical_accuracy: 0.5315 - val_top_3_accuracy: 0.7325\n\nEpoch 00018: val_loss improved from 1.93929 to 1.93120, saving model to stroke_lstm_model_weights.best.hdf5\nEpoch 19/50\n255000/255000 [==============================] - 24s 94us/step - loss: 2.1584 - categorical_accuracy: 0.4770 - top_3_accuracy: 0.6908 - val_loss: 1.8926 - val_categorical_accuracy: 0.5377 - val_top_3_accuracy: 0.7410\n\nEpoch 00019: val_loss improved from 1.93120 to 1.89257, saving model to stroke_lstm_model_weights.best.hdf5\nEpoch 20/50\n255000/255000 [==============================] - 24s 94us/step - loss: 2.1475 - categorical_accuracy: 0.4800 - top_3_accuracy: 0.6937 - val_loss: 1.9397 - val_categorical_accuracy: 0.5286 - val_top_3_accuracy: 0.7343\n\nEpoch 00020: val_loss did not improve from 1.89257\nEpoch 21/50\n255000/255000 [==============================] - 24s 94us/step - loss: 2.1365 - categorical_accuracy: 0.4821 - top_3_accuracy: 0.6963 - val_loss: 1.8808 - val_categorical_accuracy: 0.5411 - val_top_3_accuracy: 0.7424\n\nEpoch 00021: val_loss improved from 1.89257 to 1.88081, saving model to stroke_lstm_model_weights.best.hdf5\nEpoch 22/50\n255000/255000 [==============================] - 24s 94us/step - loss: 2.1286 - categorical_accuracy: 0.4842 - top_3_accuracy: 0.6973 - val_loss: 1.8723 - val_categorical_accuracy: 0.5420 - val_top_3_accuracy: 0.7464\n\nEpoch 00022: val_loss improved from 1.88081 to 1.87234, saving model to stroke_lstm_model_weights.best.hdf5\nEpoch 23/50\n255000/255000 [==============================] - 24s 94us/step - loss: 2.1150 - categorical_accuracy: 0.4867 - top_3_accuracy: 0.7003 - val_loss: 1.8742 - val_categorical_accuracy: 0.5443 - val_top_3_accuracy: 0.7471\n\nEpoch 00023: val_loss did not improve from 1.87234\nEpoch 24/50\n255000/255000 [==============================] - 24s 94us/step - loss: 2.1012 - categorical_accuracy: 0.4897 - top_3_accuracy: 0.7028 - val_loss: 1.8567 - val_categorical_accuracy: 0.5446 - val_top_3_accuracy: 0.7495\n\nEpoch 00024: val_loss improved from 1.87234 to 1.85667, saving model to stroke_lstm_model_weights.best.hdf5\nEpoch 25/50\n255000/255000 [==============================] - 24s 94us/step - loss: 2.0980 - categorical_accuracy: 0.4896 - top_3_accuracy: 0.7033 - val_loss: 1.9095 - val_categorical_accuracy: 0.5365 - val_top_3_accuracy: 0.7393\n\nEpoch 00025: val_loss did not improve from 1.85667\nEpoch 26/50\n255000/255000 [==============================] - 24s 94us/step - loss: 2.0923 - categorical_accuracy: 0.4915 - top_3_accuracy: 0.7043 - val_loss: 1.8626 - val_categorical_accuracy: 0.5478 - val_top_3_accuracy: 0.7460\n","name":"stdout"},{"output_type":"stream","text":"\nEpoch 00026: val_loss did not improve from 1.85667\nEpoch 27/50\n255000/255000 [==============================] - 24s 93us/step - loss: 2.0815 - categorical_accuracy: 0.4932 - top_3_accuracy: 0.7058 - val_loss: 1.8556 - val_categorical_accuracy: 0.5480 - val_top_3_accuracy: 0.7488\n\nEpoch 00027: val_loss improved from 1.85667 to 1.85557, saving model to stroke_lstm_model_weights.best.hdf5\nEpoch 28/50\n255000/255000 [==============================] - 24s 94us/step - loss: 2.0656 - categorical_accuracy: 0.4981 - top_3_accuracy: 0.7088 - val_loss: 1.8490 - val_categorical_accuracy: 0.5460 - val_top_3_accuracy: 0.7484\n\nEpoch 00028: val_loss improved from 1.85557 to 1.84898, saving model to stroke_lstm_model_weights.best.hdf5\nEpoch 29/50\n255000/255000 [==============================] - 24s 94us/step - loss: 2.0617 - categorical_accuracy: 0.4975 - top_3_accuracy: 0.7106 - val_loss: 1.8446 - val_categorical_accuracy: 0.5470 - val_top_3_accuracy: 0.7504\n\nEpoch 00029: val_loss improved from 1.84898 to 1.84459, saving model to stroke_lstm_model_weights.best.hdf5\nEpoch 30/50\n255000/255000 [==============================] - 24s 93us/step - loss: 2.0463 - categorical_accuracy: 0.5019 - top_3_accuracy: 0.7131 - val_loss: 1.8040 - val_categorical_accuracy: 0.5590 - val_top_3_accuracy: 0.7556\n\nEpoch 00030: val_loss improved from 1.84459 to 1.80404, saving model to stroke_lstm_model_weights.best.hdf5\nEpoch 31/50\n255000/255000 [==============================] - 24s 94us/step - loss: 2.0367 - categorical_accuracy: 0.5032 - top_3_accuracy: 0.7144 - val_loss: 1.8118 - val_categorical_accuracy: 0.5580 - val_top_3_accuracy: 0.7585\n\nEpoch 00031: val_loss did not improve from 1.80404\nEpoch 32/50\n255000/255000 [==============================] - 24s 94us/step - loss: 2.0313 - categorical_accuracy: 0.5059 - top_3_accuracy: 0.7162 - val_loss: 1.8116 - val_categorical_accuracy: 0.5578 - val_top_3_accuracy: 0.7550\n\nEpoch 00032: val_loss did not improve from 1.80404\nEpoch 33/50\n255000/255000 [==============================] - 24s 94us/step - loss: 2.0252 - categorical_accuracy: 0.5068 - top_3_accuracy: 0.7170 - val_loss: 1.7947 - val_categorical_accuracy: 0.5595 - val_top_3_accuracy: 0.7597\n\nEpoch 00033: val_loss improved from 1.80404 to 1.79465, saving model to stroke_lstm_model_weights.best.hdf5\nEpoch 34/50\n255000/255000 [==============================] - 24s 94us/step - loss: 2.0163 - categorical_accuracy: 0.5082 - top_3_accuracy: 0.7190 - val_loss: 1.7903 - val_categorical_accuracy: 0.5617 - val_top_3_accuracy: 0.7597\n\nEpoch 00034: val_loss improved from 1.79465 to 1.79035, saving model to stroke_lstm_model_weights.best.hdf5\nEpoch 35/50\n255000/255000 [==============================] - 24s 94us/step - loss: 2.0030 - categorical_accuracy: 0.5106 - top_3_accuracy: 0.7215 - val_loss: 1.8002 - val_categorical_accuracy: 0.5615 - val_top_3_accuracy: 0.7596\n\nEpoch 00035: val_loss did not improve from 1.79035\nEpoch 36/50\n255000/255000 [==============================] - 24s 93us/step - loss: 2.0003 - categorical_accuracy: 0.5122 - top_3_accuracy: 0.7218 - val_loss: 1.7809 - val_categorical_accuracy: 0.5665 - val_top_3_accuracy: 0.7619\n\nEpoch 00036: val_loss improved from 1.79035 to 1.78092, saving model to stroke_lstm_model_weights.best.hdf5\nEpoch 37/50\n255000/255000 [==============================] - 24s 94us/step - loss: 1.9907 - categorical_accuracy: 0.5141 - top_3_accuracy: 0.7248 - val_loss: 1.7742 - val_categorical_accuracy: 0.5647 - val_top_3_accuracy: 0.7621\n\nEpoch 00037: val_loss improved from 1.78092 to 1.77421, saving model to stroke_lstm_model_weights.best.hdf5\nEpoch 38/50\n255000/255000 [==============================] - 24s 94us/step - loss: 1.9810 - categorical_accuracy: 0.5166 - top_3_accuracy: 0.7257 - val_loss: 1.7741 - val_categorical_accuracy: 0.5658 - val_top_3_accuracy: 0.7644\n\nEpoch 00038: val_loss improved from 1.77421 to 1.77406, saving model to stroke_lstm_model_weights.best.hdf5\nEpoch 39/50\n255000/255000 [==============================] - 24s 94us/step - loss: 1.9783 - categorical_accuracy: 0.5172 - top_3_accuracy: 0.7258 - val_loss: 1.7945 - val_categorical_accuracy: 0.5622 - val_top_3_accuracy: 0.7595\n\nEpoch 00039: val_loss did not improve from 1.77406\nEpoch 40/50\n255000/255000 [==============================] - 24s 93us/step - loss: 1.9668 - categorical_accuracy: 0.5190 - top_3_accuracy: 0.7287 - val_loss: 1.7554 - val_categorical_accuracy: 0.5690 - val_top_3_accuracy: 0.7643\n\nEpoch 00040: val_loss improved from 1.77406 to 1.75542, saving model to stroke_lstm_model_weights.best.hdf5\nEpoch 41/50\n255000/255000 [==============================] - 24s 94us/step - loss: 1.9569 - categorical_accuracy: 0.5207 - top_3_accuracy: 0.7304 - val_loss: 1.7630 - val_categorical_accuracy: 0.5687 - val_top_3_accuracy: 0.7656\n\nEpoch 00041: val_loss did not improve from 1.75542\nEpoch 42/50\n255000/255000 [==============================] - 24s 94us/step - loss: 1.9535 - categorical_accuracy: 0.5210 - top_3_accuracy: 0.7299 - val_loss: 1.7555 - val_categorical_accuracy: 0.5685 - val_top_3_accuracy: 0.7667\n\nEpoch 00042: val_loss did not improve from 1.75542\nEpoch 43/50\n255000/255000 [==============================] - 24s 94us/step - loss: 1.9414 - categorical_accuracy: 0.5243 - top_3_accuracy: 0.7333 - val_loss: 1.7393 - val_categorical_accuracy: 0.5765 - val_top_3_accuracy: 0.7682\n\nEpoch 00043: val_loss improved from 1.75542 to 1.73927, saving model to stroke_lstm_model_weights.best.hdf5\nEpoch 44/50\n255000/255000 [==============================] - 24s 94us/step - loss: 1.9374 - categorical_accuracy: 0.5256 - top_3_accuracy: 0.7330 - val_loss: 1.7272 - val_categorical_accuracy: 0.5757 - val_top_3_accuracy: 0.7727\n\nEpoch 00044: val_loss improved from 1.73927 to 1.72720, saving model to stroke_lstm_model_weights.best.hdf5\nEpoch 45/50\n255000/255000 [==============================] - 24s 94us/step - loss: 1.9310 - categorical_accuracy: 0.5274 - top_3_accuracy: 0.7353 - val_loss: 1.7452 - val_categorical_accuracy: 0.5727 - val_top_3_accuracy: 0.7680\n\nEpoch 00045: val_loss did not improve from 1.72720\nEpoch 46/50\n255000/255000 [==============================] - 24s 94us/step - loss: 1.9249 - categorical_accuracy: 0.5270 - top_3_accuracy: 0.7364 - val_loss: 1.7294 - val_categorical_accuracy: 0.5764 - val_top_3_accuracy: 0.7707\n\nEpoch 00046: val_loss did not improve from 1.72720\nEpoch 47/50\n255000/255000 [==============================] - 24s 93us/step - loss: 1.9170 - categorical_accuracy: 0.5293 - top_3_accuracy: 0.7371 - val_loss: 1.7404 - val_categorical_accuracy: 0.5745 - val_top_3_accuracy: 0.7705\n\nEpoch 00047: val_loss did not improve from 1.72720\nEpoch 48/50\n255000/255000 [==============================] - 24s 94us/step - loss: 1.9105 - categorical_accuracy: 0.5312 - top_3_accuracy: 0.7384 - val_loss: 1.7249 - val_categorical_accuracy: 0.5770 - val_top_3_accuracy: 0.7715\n\nEpoch 00048: val_loss improved from 1.72720 to 1.72493, saving model to stroke_lstm_model_weights.best.hdf5\nEpoch 49/50\n255000/255000 [==============================] - 24s 94us/step - loss: 1.9045 - categorical_accuracy: 0.5329 - top_3_accuracy: 0.7398 - val_loss: 1.7115 - val_categorical_accuracy: 0.5818 - val_top_3_accuracy: 0.7753\n\nEpoch 00049: val_loss improved from 1.72493 to 1.71154, saving model to stroke_lstm_model_weights.best.hdf5\nEpoch 50/50\n255000/255000 [==============================] - 24s 94us/step - loss: 1.8950 - categorical_accuracy: 0.5344 - top_3_accuracy: 0.7415 - val_loss: 1.7160 - val_categorical_accuracy: 0.5772 - val_top_3_accuracy: 0.7726\n\nEpoch 00050: val_loss did not improve from 1.71154\n_________________________________________________________________\nLayer (type)                 Output Shape              Param #   \n=================================================================\nbatch_normalization_1 (Batch (None, None, 3)           12        \n_________________________________________________________________\nconv1d_1 (Conv1D)            (None, None, 48)          768       \n_________________________________________________________________\ndropout_1 (Dropout)          (None, None, 48)          0         \n_________________________________________________________________\nconv1d_2 (Conv1D)            (None, None, 64)          15424     \n_________________________________________________________________\ndropout_2 (Dropout)          (None, None, 64)          0         \n_________________________________________________________________\nconv1d_3 (Conv1D)            (None, None, 96)          18528     \n_________________________________________________________________\ndropout_3 (Dropout)          (None, None, 96)          0         \n_________________________________________________________________\ncu_dnnlstm_1 (CuDNNLSTM)     (None, None, 128)         115712    \n_________________________________________________________________\ndropout_4 (Dropout)          (None, None, 128)         0         \n_________________________________________________________________\ncu_dnnlstm_2 (CuDNNLSTM)     (None, 128)               132096    \n_________________________________________________________________\ndropout_5 (Dropout)          (None, 128)               0         \n_________________________________________________________________\ndense_1 (Dense)              (None, 512)               66048     \n_________________________________________________________________\ndropout_6 (Dropout)          (None, 512)               0         \n_________________________________________________________________\ndense_2 (Dense)              (None, 340)               174420    \n=================================================================\nTotal params: 523,008\nTrainable params: 523,002\nNon-trainable params: 6\n_________________________________________________________________\n","name":"stdout"}]},{"metadata":{"trusted":true},"cell_type":"code","source":"stroke_read_model.load_weights(weight_path)\nlstm_results = stroke_read_model.evaluate(test_X, test_y, batch_size = 4096)\nprint('Accuracy: %2.1f%%, Top 3 Accuracy %2.1f%%' % (100*lstm_results[1], 100*lstm_results[2]))","execution_count":66,"outputs":[{"output_type":"stream","text":"17000/17000 [==============================] - 1s 34us/step\nAccuracy: 46.5%, Top 3 Accuracy 68.1%\n","name":"stdout"}]},{"metadata":{"trusted":true},"cell_type":"code","source":"sub_df = pd.read_csv(test_path)\nsub_df['drawing'] = sub_df['drawing'].map(_stack_it)","execution_count":72,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"sub_vec = np.stack(sub_df['drawing'].values, 0)\nsub_pred = stroke_read_model.predict(sub_vec, verbose=True, batch_size=4096)","execution_count":73,"outputs":[{"output_type":"stream","text":"112199/112199 [==============================] - 4s 34us/step\n","name":"stdout"}]},{"metadata":{"trusted":true},"cell_type":"code","source":"top_3_pred = [word_encoder.classes_[np.argsort(-1*c_pred)[:3]] for c_pred in sub_pred]","execution_count":74,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"top_3_pred = [' '.join([col.replace(' ', '_') for col in row]) for row in top_3_pred]\ntop_3_pred[:3]","execution_count":75,"outputs":[{"output_type":"execute_result","execution_count":75,"data":{"text/plain":"['radio stereo stove',\n 'hamburger hockey_puck bottlecap',\n 'castle The_Great_Wall_of_China crown']"},"metadata":{}}]},{"metadata":{"trusted":true},"cell_type":"code","source":"sub_df['word'] = top_3_pred\nsub_df[['key_id', 'word']].to_csv('submission.csv', index=False)","execution_count":76,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"","execution_count":null,"outputs":[]}],"metadata":{"kernelspec":{"display_name":"Python 3","language":"python","name":"python3"},"language_info":{"name":"python","version":"3.6.4","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"}},"nbformat":4,"nbformat_minor":1}