In [1]:

    

import numpy as np
import matplotlib.pyplot as plt
%matplotlib inline
from keras.utils import np_utils
from keras.preprocessing.image import ImageDataGenerator
from keras.models import Sequential
from keras.layers import Dense, Dropout, Activation, Flatten
from keras.layers import Convolution2D, MaxPooling2D, AveragePooling2D
from keras.optimizers import Adam
import glob
from PIL import Image

import keras
from keras.preprocessing.image import ImageDataGenerator
from keras.models import Sequential
from keras.layers import Dense, Dropout, Activation, Flatten
from keras.layers import Conv2D, MaxPooling2D
from keras.utils import np_utils
from keras.layers.core import Flatten, Dense, Dropout, Lambda


    

    




Using TensorFlow backend.

In [2]:

    

from keras.preprocessing import image

BATCH_SIZE = 64
PATH="data/"

def get_fit_sample():
    gen = image.ImageDataGenerator()
    sample_batches = gen.flow_from_directory(PATH+'valid', target_size=(224,224), 
            class_mode='categorical', shuffle=False, batch_size=1000)
    imgs, labels = next(sample_batches)
    return imgs

gen = image.ImageDataGenerator(featurewise_std_normalization=True)
gen.fit(get_fit_sample())
val_batches = gen.flow_from_directory(PATH+'valid', target_size=(224,224), 
            class_mode='categorical', shuffle=True, batch_size=BATCH_SIZE)

gen = image.ImageDataGenerator(featurewise_std_normalization=True)
gen.fit(get_fit_sample())
batches = gen.flow_from_directory(PATH+'train', target_size=(224,224), 
            class_mode='categorical', shuffle=True, batch_size=BATCH_SIZE)

# from utils import plots
#imgs,labels = next(batches)
#plots(imgs[:2])


    

    




Found 1000 images belonging to 2 classes.
Found 1000 images belonging to 2 classes.
Found 1000 images belonging to 2 classes.
Found 24000 images belonging to 2 classes.

In [ ]:

    




    

In [3]:

    

CLASSES = 2
INPUT_SHAPE = (224,224,3)
model = Sequential()
    
# Block 1
model.add(Conv2D(64, (3, 3), activation='relu', padding='same', name='block1_conv1', input_shape=INPUT_SHAPE))
model.add(Conv2D(64, (3, 3), activation='relu', padding='same', name='block1_conv2'))
model.add(MaxPooling2D((2, 2), strides=(2, 2), name='block1_pool'))

# Block 2
model.add(Conv2D(128, (3, 3), activation='relu', padding='same', name='block2_conv1'))
model.add(Conv2D(128, (3, 3), activation='relu', padding='same', name='block2_conv2'))
model.add(MaxPooling2D((2, 2), strides=(2, 2), name='block2_pool'))

# Block 3
model.add(Conv2D(256, (3, 3), activation='relu', padding='same', name='block3_conv1'))
model.add(Conv2D(256, (3, 3), activation='relu', padding='same', name='block3_conv2'))
model.add(Conv2D(256, (3, 3), activation='relu', padding='same', name='block3_conv3'))
model.add(MaxPooling2D((2, 2), strides=(2, 2), name='block3_pool'))

# Block 4
model.add(Conv2D(512, (3, 3), activation='relu', padding='same', name='block4_conv1'))
model.add(Conv2D(512, (3, 3), activation='relu', padding='same', name='block4_conv2'))
model.add(Conv2D(512, (3, 3), activation='relu', padding='same', name='block4_conv3'))
model.add(MaxPooling2D((2, 2), strides=(2, 2), name='block4_pool'))

# Block 5
model.add(Conv2D(512, (3, 3), activation='relu', padding='same', name='block5_conv1'))
model.add(Conv2D(512, (3, 3), activation='relu', padding='same', name='block5_conv2'))
model.add(Conv2D(512, (3, 3), activation='relu', padding='same', name='block5_conv3'))
model.add(MaxPooling2D((2, 2), strides=(2, 2), name='block5_pool'))

# Classification block
model.add(Flatten(name='flatten'))
model.add(Dense(4096, activation='relu', name='fc1'))
model.add(Dropout(0.5))
model.add(Dense(4096, activation='relu', name='fc2'))
model.add(Dropout(0.5))
model.add(Dense(CLASSES, activation='softmax', name='predictions'))

from keras.optimizers import SGD
sgd = SGD(lr=0.01, decay=0.0005, momentum=0.9, nesterov=False)


model.compile(optimizer=sgd, loss='mean_squared_error', metrics=['accuracy'])


    

In [4]:

    

%%time
hist = model.fit_generator(batches,  steps_per_epoch=1000, epochs=25, validation_data=val_batches, validation_steps=1000)
 
model.save('ConvNet-D-vgg16-large-dataset.h5')

# http://qiita.com/TypeNULL/items/4e4d7de11ab4361d6085
loss = hist.history['loss']
val_loss = hist.history['val_loss']
nb_epoch = len(loss)
plt.plot(range(nb_epoch), loss, marker='.', label='loss')
plt.plot(range(nb_epoch), val_loss, marker='.', label='val_loss')
plt.legend(loc='best', fontsize=10)
plt.grid()
plt.xlabel('epoch')
plt.ylabel('loss')
plt.show()


    

    




Epoch 1/25
1000/1000 [==============================] - 1378s - loss: 0.2352 - acc: 0.5810 - val_loss: 0.2127 - val_acc: 0.6471
Epoch 2/25
1000/1000 [==============================] - 1374s - loss: 0.1946 - acc: 0.6914 - val_loss: 0.2010 - val_acc: 0.6869
Epoch 3/25
1000/1000 [==============================] - 1374s - loss: 0.1592 - acc: 0.7659 - val_loss: 0.1581 - val_acc: 0.7571
Epoch 4/25
1000/1000 [==============================] - 1374s - loss: 0.1262 - acc: 0.8219 - val_loss: 0.1135 - val_acc: 0.8469
Epoch 5/25
1000/1000 [==============================] - 1374s - loss: 0.1007 - acc: 0.8608 - val_loss: 0.0949 - val_acc: 0.8600
Epoch 6/25
1000/1000 [==============================] - 1375s - loss: 0.0816 - acc: 0.8896 - val_loss: 0.0770 - val_acc: 0.8990
Epoch 7/25
1000/1000 [==============================] - 1374s - loss: 0.0681 - acc: 0.9092 - val_loss: 0.0699 - val_acc: 0.9020
Epoch 8/25
1000/1000 [==============================] - 1373s - loss: 0.0590 - acc: 0.9220 - val_loss: 0.0612 - val_acc: 0.9160
Epoch 9/25
1000/1000 [==============================] - 1375s - loss: 0.0501 - acc: 0.9351 - val_loss: 0.0624 - val_acc: 0.9191
Epoch 10/25
1000/1000 [==============================] - 1375s - loss: 0.0426 - acc: 0.9464 - val_loss: 0.0583 - val_acc: 0.9251
Epoch 11/25
1000/1000 [==============================] - 1374s - loss: 0.0360 - acc: 0.9549 - val_loss: 0.0569 - val_acc: 0.9280
Epoch 12/25
1000/1000 [==============================] - 1378s - loss: 0.0309 - acc: 0.9616 - val_loss: 0.0566 - val_acc: 0.9210
Epoch 13/25
1000/1000 [==============================] - 1381s - loss: 0.0252 - acc: 0.9694 - val_loss: 0.0517 - val_acc: 0.9300
Epoch 14/25
1000/1000 [==============================] - 1380s - loss: 0.0210 - acc: 0.9751 - val_loss: 0.0529 - val_acc: 0.9280
Epoch 15/25
1000/1000 [==============================] - 1379s - loss: 0.0181 - acc: 0.9792 - val_loss: 0.0524 - val_acc: 0.9310
Epoch 16/25
1000/1000 [==============================] - 1380s - loss: 0.0138 - acc: 0.9851 - val_loss: 0.0505 - val_acc: 0.9431
Epoch 17/25
1000/1000 [==============================] - 1380s - loss: 0.0125 - acc: 0.9861 - val_loss: 0.0543 - val_acc: 0.9341
Epoch 18/25
1000/1000 [==============================] - 1380s - loss: 0.0101 - acc: 0.9890 - val_loss: 0.0521 - val_acc: 0.9340
Epoch 19/25
1000/1000 [==============================] - 1381s - loss: 0.0089 - acc: 0.9905 - val_loss: 0.0569 - val_acc: 0.9330
Epoch 20/25
1000/1000 [==============================] - 1380s - loss: 0.0076 - acc: 0.9919 - val_loss: 0.0577 - val_acc: 0.9311
Epoch 21/25
1000/1000 [==============================] - 1380s - loss: 0.0063 - acc: 0.9939 - val_loss: 0.0532 - val_acc: 0.9310
Epoch 22/25
1000/1000 [==============================] - 1380s - loss: 0.0058 - acc: 0.9943 - val_loss: 0.0604 - val_acc: 0.9249
Epoch 23/25
1000/1000 [==============================] - 1380s - loss: 0.0053 - acc: 0.9948 - val_loss: 0.0566 - val_acc: 0.9310
Epoch 24/25
1000/1000 [==============================] - 1380s - loss: 0.0050 - acc: 0.9951 - val_loss: 0.0585 - val_acc: 0.9309
Epoch 25/25
1000/1000 [==============================] - 1380s - loss: 0.0045 - acc: 0.9956 - val_loss: 0.0560 - val_acc: 0.9299






    













    




CPU times: user 6h 6min 41s, sys: 55min 54s, total: 7h 2min 35s
Wall time: 9h 34min 11s

In [25]:

    

%%time
# Prediction

gen = image.ImageDataGenerator(featurewise_std_normalization=True)
gen.fit(get_fit_sample())
test_batches = gen.flow_from_directory(PATH+'test', target_size=(224,224), 
           class_mode=None, shuffle=False, batch_size=BATCH_SIZE)
print(test_batches.samples)

preds = model.predict_generator(test_batches, test_batches.samples / BATCH_SIZE, verbose=1)


    

    




Found 1000 images belonging to 2 classes.
Found 12500 images belonging to 1 classes.
12500
196/195 [==============================] - 59s    
CPU times: user 29.9 s, sys: 2.25 s, total: 32.1 s
Wall time: 1min 2s

In [35]:

    

import bcolz
#https://github.com/fastai/courses/blob/master/deeplearning1/nbs/utils.py#L175
def save_array(fname, arr):
    c=bcolz.carray(arr, rootdir=fname, mode='w')
    c.flush()
    
print(preds.shape)
filenames = test_batches.filenames
print(filenames[:5])
%mkdir -p data/results
save_array(PATH  + 'results/test_preds.dat', preds)
save_array(PATH + 'results/filenames.dat', filenames)


    

    




(12500, 2)
['unknown/4049.jpg', 'unknown/11042.jpg', 'unknown/769.jpg', 'unknown/8045.jpg', 'unknown/3315.jpg']

In [68]:

    

isdog = preds[:,1]

print("Raw Predictions: " + str(our_predictions[:5]))
print("Raw Predictions: " + str(isdog[:5]))
print("Mid Predictions: " + str(len(isdog[(isdog < .6) & (isdog > .4)])))
print("Edge Predictions: " + str(len(isdog[(isdog == 1) | (isdog == 0)])))


    

    




Raw Predictions: [  7.32483896e-09   1.00000000e+00   3.00929503e-04   1.09451648e-08
   3.29333858e-08]
Raw Predictions: [  1.00000000e+00   1.72287010e-14   9.99699116e-01   1.00000000e+00
   1.00000000e+00]
Mid Predictions: 128
Edge Predictions: 940

In [86]:

    

#https://github.com/fastai/courses/blob/master/deeplearning1/nbs/dogs_cats_redux.ipynb
#from utils import plots
def plots(ims, figsize=(12,6), rows=1, interp=False, titles=None):
    if type(ims[0]) is np.ndarray:
        ims = np.array(ims).astype(np.uint8)
        if (ims.shape[-1] != 3):
            ims = ims.transpose((0,2,3,1))
    f = plt.figure(figsize=figsize)
    for i in range(len(ims)):
        sp = f.add_subplot(rows, len(ims)//rows, i+1)
        sp.axis('Off')
        if titles is not None:
            sp.set_title(titles[i], fontsize=16)
        plt.imshow(ims[i], interpolation=None if interp else 'none')
def plots_idx(idx, titles=None):
    plots([image.load_img("data/test/" + filenames[i]) for i in idx], titles=titles)

isdog = preds[:,1]

for i in range(0, 100, 5):
  plots_idx(list(range(i, i + 5)), ["dog %1.3f"%p if p >= 0.5 else "cat %1.3f"%(1-p) for p in isdog[i:i+5] ])


    

In [93]:

    

isdog_clipped = isdog.clip(min=0.05, max=0.95)
ids = np.array([int(f[8:f.find('.')]) for f in filenames])
subm = np.stack([ids,isdog_clipped], axis=1)
subm[:5]


    

    
Out[93]:





array([[  4.04900000e+03,   9.49999988e-01],
       [  1.10420000e+04,   5.00000007e-02],
       [  7.69000000e+02,   9.49999988e-01],
       [  8.04500000e+03,   9.49999988e-01],
       [  3.31500000e+03,   9.49999988e-01]])

In [94]:

    

from IPython.display import FileLink

np.savetxt(PATH + "results/submission2.csv", subm, fmt='%d,%.5f', header='id,label', comments='')
FileLink(PATH + "results/submission2.csv")


    

    
Out[94]:




data/results/submission2.csv

In [ ]: