tokenize bottlenecks

tokenize.py

#docker run -d -p 8080:8080 -v /root:/data   gcr.io/deeplearning-platform-release/tf-cpu.1-13
import os
import sys
import tempfile
import shutil
import numpy as np
np.set_printoptions(threshold=np.inf)
from keras.preprocessing.image import ImageDataGenerator, img_to_array, load_img
from keras.models import Sequential
from keras.layers import Dropout, Flatten, Dense
from keras import applications
from keras.utils.np_utils import to_categorical
import matplotlib.pyplot as plt
import math
import cv2

# dimensions of our images.
img_width, img_height = 224, 224

top_model_weights_path = 'bottleneck_fc_model.h5'

def save_bottlebeck_features(model, train_data_dir):
    datagen = ImageDataGenerator()#rescale=1. / 255)
    generator = datagen.flow_from_directory(
        train_data_dir,
        target_size=(img_width, img_height),
        batch_size=1,
        class_mode=None,
        shuffle=False)
    bottleneck_features_train = model.predict_generator(generator, 1)
    return bottleneck_features_train

model = applications.xception.Xception(include_top=False, weights='imagenet')

d = sys.argv[1]
for f in os.listdir(d):
    path0 = tempfile.mkdtemp()
    path1 = "%s/%s" % (path0, "1")
    os.mkdir(path1)
    shutil.copyfile("%s/%s"%(d,f),"%s/%s"%(path1,f))
    a = save_bottlebeck_features(model,path0)
    aflat = np.multiply(np.divide(a,np.amax(a)),8).astype('uint8').flatten()
    i = 0
    sys.stdout.write("%s\t" % f)
    for x in np.nditer(aflat):
        if x > 0:
            for k in range(0,x):
                sys.stdout.write('X%X ' % i)
            i = i + 1
    shutil.rmtree(path0)

#np.save('bottleneck_features_train.npy', a)