In [1]:

    

import numpy as np
from PIL import Image
import matplotlib.pyplot as plt
import time
%matplotlib inline
import lmdb
from caffe.proto import caffe_pb2
# import caffe


    

In [72]:

    

train = lmdb.open('./data/48x48/ndsb_train_lmdb/')

image_dims = (48, 48)
with train.begin() as txn:
    c = txn.cursor()
    k0, v0 = iter(c).next()
    
datum = caffe_pb2.Datum()
def bs_to_l(bs):
    datum.ParseFromString(bs)
    return dd.label

def bs_to_im(bs):
    datum.ParseFromString(bs)
    image_dims = (datum.height, datum.width)
    im = np.array(Image.frombytes('L', image_dims, datum.data))[:, :, None]
    return im
print bs_to_l(v0)

# plt.imshow(np.squeeze(bs_to_im(v0)), cmap='gray', interpolation='none')


    

    




0

In [66]:

    

%%timeit
q = bs_to_im(v0)


    

    




10000 loops, best of 3: 72.9 µs per loop

In [73]:

    

tic = time.time()
test = lmdb.open('/data/ndsb/ndsb_test_lmdb/', readonly=True)

with test.begin() as txn:
    c = txn.cursor()
    zz = [(k.split('_')[1], bs_to_im(v), bs_to_l(v)) 
          for k, v in c]
len(zz)
print time.time() - tic


    

    




15.9699380398

In [71]:

    

zz[0]


    

    
Out[71]:





('/data/ndsb/test/68720.jpg', array([[[255],
         [255],
         [255],
         ..., 
         [255],
         [255],
         [255]],
 
        [[255],
         [255],
         [255],
         ..., 
         [255],
         [255],
         [255]],
 
        [[255],
         [255],
         [255],
         ..., 
         [255],
         [255],
         [255]],
 
        ..., 
        [[255],
         [255],
         [255],
         ..., 
         [255],
         [255],
         [255]],
 
        [[255],
         [255],
         [255],
         ..., 
         [255],
         [255],
         [255]],
 
        [[255],
         [255],
         [255],
         ..., 
         [255],
         [255],
         [255]]], dtype=uint8), 0)

In [34]:

    

plt.imshow(img, cmap='gray')


    

    
Out[34]:





<matplotlib.image.AxesImage at 0x7efff4512610>






    

In [36]:

    

# Scale to standardize input dimensions.
input_ = np.zeros((len(inputs),
    image_dims[0], image_dims[1], inputs[0].shape[2]),
    dtype=np.float32)
for ix, in_ in enumerate(inputs):
    input_[ix] = caffe.io.resize_image(in_, image_dims)

if oversample:
    # Generate center, corner, and mirrored crops.
    input_ = caffe.io.oversample(input_, crop_dims)
else:
    # Take center crop.
    center = np.array(self.image_dims) / 2.0
    crop = np.tile(center, (1, 2))[0] + np.concatenate([
        -self.crop_dims / 2.0,
        self.crop_dims / 2.0
    ])
    input_ = input_[:, crop[0]:crop[2], crop[1]:crop[3], :]


    

    




---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
<ipython-input-36-7d5d4879fadc> in <module>()
      6     input_[ix] = caffe.io.resize_image(in_, image_dims)
      7 
----> 8 if oversample:
      9     # Generate center, corner, and mirrored crops.
     10     input_ = caffe.io.oversample(input_, crop_dims)

NameError: name 'oversample' is not defined

In [20]:

    

datum = caffe_pb2.Datum()
datum.ParseFromString(v0)
# print datum.label

image_dims = (datum.height, datum.width)
img = Image.frombytes('L', image_dims, datum.data)
plt.imshow(img, cmap='gray', interpolation='none')


    

    
Out[20]:





<matplotlib.image.AxesImage at 0x7f9d388e5310>






    

In [80]:

    

import tools.my_io as my_io

tic = time.time()
qq = my_io.load_lmdb('./data/48x48/ndsb_train_lmdb/')
print 'Load db time:', time.time()-tic


    

    




Load db time: 2.4930639267

In [85]:

    

p, im, l = qq[0]
p
im
l


    

    
Out[85]:





0