In [3]:

    

%matplotlib inline


    

In [11]:

    

##For sample data
path = "data/distracted-driving/sample/"
results_path = 'distracted-driving-results/sample'

##For full data
#path = "data/distracted-driving/"
#results_path = 'distracted-driving-results/sample'


    

In [5]:

    

from __future__ import division,print_function

import os, json
from glob import glob
import numpy as np
np.set_printoptions(precision=4, linewidth=100)
from matplotlib import pyplot as plt


    

In [6]:

    

import utils; reload(utils)
from utils import plots


    

    




Using gpu device 0: Tesla K80 (CNMeM is disabled, cuDNN 5103)
/home/ubuntu/anaconda2/lib/python2.7/site-packages/theano/sandbox/cuda/__init__.py:600: UserWarning: Your cuDNN version is more recent than the one Theano officially supports. If you see any problems, try updating Theano or downgrading cuDNN to version 5.
  warnings.warn(warn)
Using Theano backend.

In [6]:

    

batch_size=64


    

In [7]:

    

# Import our class, and instantiate
import vgg16; reload(vgg16)
from vgg16 import Vgg16


    

In [8]:

    

vgg = Vgg16()


    

In [9]:

    

batches = vgg.get_batches(path+'train', batch_size=4)


    

    




Found 19487 images belonging to 10 classes.

In [10]:

    

imgs,labels = next(batches)
plots(imgs, titles=labels)


    

In [11]:

    

vgg.predict(imgs, True)


    

    
Out[11]:





(array([ 0.8783,  0.4636,  0.4592,  0.5569], dtype=float32),
 array([785, 785, 656, 785]),
 [u'seat_belt', u'seat_belt', u'minivan', u'seat_belt'])

In [12]:

    

vgg.finetune(batches)


    

In [13]:

    

val_batches = vgg.get_batches(path+'valid', batch_size=batch_size)


    

    




Found 2237 images belonging to 10 classes.

In [14]:

    

vgg.fit(batches, val_batches, nb_epoch=1)


    

    




Epoch 1/1
19487/19487 [==============================] - 564s - loss: 2.4660 - acc: 0.5121 - val_loss: 0.5290 - val_acc: 0.8418

In [15]:

    

vgg.predict(imgs, True)


    

    
Out[15]:





(array([ 0.9858,  0.9999,  0.4853,  0.9999], dtype=float32),
 array([4, 4, 4, 5]),
 ['c4', 'c4', 'c4', 'c5'])

In [17]:

    

batches, preds = vgg.test(path+'test', batch_size = 32)


    

    




Found 79726 images belonging to 1 classes.

In [18]:

    

#Save our test results arrays so we can use them again later
filenames = batches.filenames
np.save(path + 'test_preds.dat', preds)
np.save(path + 'filenames.dat', filenames)


    

In [12]:

    

#Load our test predictions from file
preds = np.load(path + '/test_preds.dat.npy')
filenames = np.load(path + '/filenames.dat.npy')


    

In [13]:

    

preds


    

    
Out[13]:





array([[  1.6076e-02,   9.0095e-06,   2.0508e-10, ...,   2.6182e-05,   1.1057e-05,   9.1267e-01],
       [  5.5489e-01,   4.4217e-02,   1.8653e-08, ...,   8.3028e-02,   2.9943e-01,   6.5743e-03],
       [  1.3442e-05,   6.7272e-08,   5.2875e-06, ...,   9.6147e-10,   6.5628e-01,   2.4434e-06],
       ..., 
       [  2.4027e-02,   1.2791e-04,   5.7995e-10, ...,   1.6099e-01,   5.8294e-01,   1.3516e-07],
       [  9.2921e-01,   5.6042e-06,   2.1042e-07, ...,   9.1156e-04,   3.5594e-04,   6.1193e-04],
       [  7.9547e-01,   5.5362e-06,   5.1936e-09, ...,   3.9074e-07,   4.3125e-05,   1.2740e-03]], dtype=float32)

In [20]:

    

file_ids = np.array([f[7:] for f in filenames])
file_ids.T


    

    
Out[20]:





array(['img_81601.jpg', 'img_14887.jpg', 'img_62885.jpg', ..., 'img_98750.jpg', 'img_42858.jpg',
       'img_98905.jpg'], 
      dtype='|S14')

In [18]:

    

import pandas as pd

df = pd.DataFrame(data=preds, columns=['c0','c1','c2','c3','c4','c5','c6','c7','c8','c9'])
df.head()


    

    
Out[18]:






  

    

      

      
c0
      
c1
      
c2
      
c3
      
c4
      
c5
      
c6
      
c7
      
c8
      
c9
    
  
  

    

      
0
      
1.607646e-02
      
9.009521e-06
      
2.050786e-10
      
1.484431e-05
      
0.000071
      
4.667617e-04
      
0.070657
      
2.618243e-05
      
0.000011
      
0.912667
    
    

      
1
      
5.548896e-01
      
4.421740e-02
      
1.865346e-08
      
2.454511e-04
      
0.000436
      
1.922752e-03
      
0.009259
      
8.302806e-02
      
0.299428
      
0.006574
    
    

      
2
      
1.344189e-05
      
6.727246e-08
      
5.287497e-06
      
1.516627e-03
      
0.322349
      
1.065140e-05
      
0.019819
      
9.614698e-10
      
0.656283
      
0.000002
    
    

      
3
      
1.064366e-07
      
1.235221e-05
      
5.170768e-04
      
1.782592e-07
      
0.001629
      
2.265113e-08
      
0.006913
      
1.562694e-05
      
0.990537
      
0.000376
    
    

      
4
      
6.393413e-02
      
5.363804e-03
      
1.834831e-05
      
7.161833e-04
      
0.000056
      
1.169559e-03
      
0.729950
      
1.341563e-02
      
0.171388
      
0.013988
    
  

In [24]:

    

filename_series = pd.Series(file_ids)
df['img'] = filename_series
cols = df.columns.tolist()
cols = cols[-1:] + cols[:-1]
df = df[cols] 
df.head()


    

    
Out[24]:






  

    

      

      
img
      
c0
      
c1
      
c2
      
c3
      
c4
      
c5
      
c6
      
c7
      
c8
      
c9
    
  
  

    

      
0
      
img_81601.jpg
      
1.607646e-02
      
9.009521e-06
      
2.050786e-10
      
1.484431e-05
      
0.000071
      
4.667617e-04
      
0.070657
      
2.618243e-05
      
0.000011
      
0.912667
    
    

      
1
      
img_14887.jpg
      
5.548896e-01
      
4.421740e-02
      
1.865346e-08
      
2.454511e-04
      
0.000436
      
1.922752e-03
      
0.009259
      
8.302806e-02
      
0.299428
      
0.006574
    
    

      
2
      
img_62885.jpg
      
1.344189e-05
      
6.727246e-08
      
5.287497e-06
      
1.516627e-03
      
0.322349
      
1.065140e-05
      
0.019819
      
9.614698e-10
      
0.656283
      
0.000002
    
    

      
3
      
img_45125.jpg
      
1.064366e-07
      
1.235221e-05
      
5.170768e-04
      
1.782592e-07
      
0.001629
      
2.265113e-08
      
0.006913
      
1.562694e-05
      
0.990537
      
0.000376
    
    

      
4
      
img_22633.jpg
      
6.393413e-02
      
5.363804e-03
      
1.834831e-05
      
7.161833e-04
      
0.000056
      
1.169559e-03
      
0.729950
      
1.341563e-02
      
0.171388
      
0.013988
    
  

In [26]:

    

df.to_csv('submission.csv', index=False)


    

In [ ]: