In [1]:

    

import os
import pandas as pd
import numpy as np
%matplotlib inline
from matplotlib import pylab as plt
import cPickle as pickle
import itertools
from matplotlib.path import Path
import matplotlib.patches as patches


    

In [2]:

    

from fuel.datasets.hdf5 import H5PYDataset
import fuel
datasource = 'sketch'
datasource_dir = os.path.join(fuel.config.data_path, datasource)
datasource_fname = os.path.join(datasource_dir , datasource+'.hdf5')


    

    




Using gpu device 0: GeForce GT 650M

In [3]:

    

!mkdir -p {datasource_dir}


    

In [4]:

    

import scipy.io
import os

m = scipy.io.loadmat(os.path.join(datasource_dir,'sketches_matlab/sketches.mat'))


    

In [5]:

    

sketches = m['D']
sketches.shape


    

    
Out[5]:





(20000, 3)

In [6]:

    

def imgpoints(img):
    points = []
    
    strikes = sketches[img,2][:,0]
    lastx = lasty = 0
    for i in range(5,len(strikes)-3):
        s = strikes[i][0]
        j = s.find('d="M')
        assert j >=0
        s = s[j+len('d="M'):]
        assert s.endswith('"/>')
        s = s[:-len('"/>')]
        s = s.replace('C',' ').replace('L',' ').replace(',',' ')
        s = s.split()
        s = map(float,s)
        assert len(s) % 2 == 0
        count = 0
        for j in range(0,len(s),2):
            x, y = s[j],s[j+1]
            if np.isnan(x) or np.isnan(y):
                assert np.isnan(x) and np.isnan(y)
                continue
            points.append((x-lastx,y-lasty,count == 0))
            count += 1
            lastx,lasty = x,y
        assert count > 1
    return np.array(points[1:])


    

In [7]:

    

sketches_arrays = []
for img in xrange(len(sketches)):
    sketch = imgpoints(img)[1:]
    sketches_arrays.append(sketch)


    

In [8]:

    

lens = map(len, sketches_arrays)


    

In [9]:

    

min(lens), max(lens)


    

    
Out[9]:





(17, 3959)

In [10]:

    

plt.hist(lens,bins=50)
plt.xlabel('#points')
plt.ylabel('#sketches')
plt.title('Distribution of number of points in a sketch');


    

In [11]:

    

sketches_arrays_idx = filter(lambda i: len(sketches_arrays[i])<1200, range(len(sketches_arrays)))
N = len(sketches_arrays_idx)
N


    

    
Out[11]:





19885

In [12]:

    

xys = np.vstack([sketches_arrays[i][:,:2] for i in sketches_arrays_idx])
xys.shape


    

    
Out[12]:





(5693675, 2)

In [13]:

    

s = xys.std()
s


    

    
Out[13]:





22.718561804436927

In [14]:

    

X = np.empty((N,1200,3))
for i,j in enumerate(sketches_arrays_idx):
    p = sketches_arrays[j]
    n = len(p)
    X[i,:n,:2] = p[:,:2]/s
    X[i,n:,:2] = 0
    X[i,:n,2] =  p[:,2]
    X[i,n:,2] = 1


    

In [15]:

    

X.shape


    

    
Out[15]:





(19885, 1200, 3)

In [16]:

    

import random
X = X[random.sample(xrange(N),N)]


    

In [17]:

    

!rm {datasource_fname}


    

In [18]:

    

import h5py
fp = h5py.File(datasource_fname, mode='w')
image_features = fp.create_dataset('features', X.shape, dtype='float32')


    

In [19]:

    

image_features[...] = X


    

In [20]:

    

N_train = int(9*len(X)/10)
N_train


    

    
Out[20]:





17896

In [21]:

    

split_dict = {
    'train': {'features': (0, N_train)},
    'test': {'features': (N_train, N)}
}
fp.attrs['split'] = H5PYDataset.create_split_array(split_dict)


    

In [22]:

    

fp.flush()
fp.close()


    

In [24]:

    

train_set = H5PYDataset(datasource_fname, which_sets=['train'])


    

In [25]:

    

train_set.num_examples


    

    
Out[25]:





17896

In [26]:

    

handle = train_set.open()
data = train_set.get_data(handle, slice(0, 16))
data[0].shape, data[0].dtype, data[0].max()


    

    
Out[26]:





((16, 1200, 3), dtype('float32'), 13.689247)

In [27]:

    

train_set.close(handle)


    

In [28]:

    

from sketch import drawpoints


    

    




/Users/udi/anaconda/lib/python2.7/site-packages/matplotlib/__init__.py:1318: UserWarning:  This call to matplotlib.use() has no effect
because the backend has already been chosen;
matplotlib.use() must be called *before* pylab, matplotlib.pyplot,
or matplotlib.backends is imported for the first time.

  warnings.warn(_use_error_msg)

In [29]:

    

plt.figure(figsize=(10,10))
count = 1
M = 4
for points in data[0]:
    plt.subplot(2,2,count)
    count += 1
    drawpoints(points)
    if count > M: break


    

In [ ]: