

In [84]:

    
%matplotlib inline
import matplotlib.pyplot as plt
import numpy as np



In [5]:

    
DATA = '../data/'



In [8]:

    
from scipy.misc import imsave, fromimage, toimage
from PIL import Image, ImageOps

WIDTH = HEIGHT = 224
def load_and_crop_image(filename, target_size):
    return ImageOps.fit(Image.open(filename), target_size)

Carga imágenes



In [15]:

    
gatos = np.array([fromimage(load_and_crop_image(DATA+'/gatos/cat.{:04}.jpg'.format(i), (WIDTH, HEIGHT))) for i in range(100)])
perros = np.array([fromimage(load_and_crop_image(DATA+'/gatos/dog.{:04}.jpg'.format(i), (WIDTH, HEIGHT))) for i in range(100)])

imágenes = np.concatenate((gatos, perros), axis=0)
labels = np.zeros(len(gatos) + len(perros), dtype=int)
labels[:len(gatos)] = 1



In [141]:

    
toimage(imágenes[0])









    Out[141]:

Dividirlas en grupos



In [13]:

    
características = np.load(DATA+'/feats/all_feats.npy')
indices1 = np.load(DATA+'/feats/indices1.npy')
indices2 = np.load(DATA+'/feats/indices2.npy')



In [ ]:

    
imágenes1 = imágenes[indices1]
car1 = características[indices1]
imágenes2 = imágenes[indices2]
car2 = características[indices2]



In [122]:

    
len(imágenes1)









    Out[122]:





10



In [123]:

    
toimage(imágenes1[0])









    Out[123]:



In [ ]:



In [ ]:



In [ ]:



In [ ]:



In [69]:

    
%run vgg16.py









    



Using TensorFlow backend.






    



Model loaded.



In [77]:

    
all_cats









    Out[77]:





[281, 282, 283, 284, 285, 286, 287]



In [145]:

    
preds[all_cats].sum()









    Out[145]:





0.0013668686412984243



In [153]:

    
preds = model.predict(img_to_vgg_input(imágenes1[5]))[0]



In [152]:

    
[synsets[idx] for idx in np.argsort(preds)[::-1][:5]]









    Out[152]:





[['n02123045', 'tabby, tabby cat'],
 ['n02124075', 'Egyptian cat'],
 ['n02123159', 'tiger cat'],
 ['n02123394', 'Persian cat'],
 ['n02127052', 'lynx, catamount']]



In [ ]: