annotate pylearn/datasets/flickr.py @ 601:fd95ff96dd47

updated flickr to row-major files
author James Bergstra <bergstrj@iro.umontreal.ca>
date Wed, 14 Jan 2009 17:00:57 -0500
parents e56303df3c77
children 28f7dc848efc
rev   line source
600
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
1 """
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
2 Routines to load variations on the Flickr image dataset.
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
3 """
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
4 from __future__ import absolute_import
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
5
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
6 import os
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
7 import numpy
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
8
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
9 from ..io import filetensor
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
10 from .config import data_root
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
11 from .dataset import Dataset
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
12
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
13
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
14 def test_10class():
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
15 #TODO: make path an option,
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
16 #TODO: make default path relative to data_root()
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
17 f = open('flickr_10classes_test.ft')
601
fd95ff96dd47 updated flickr to row-major files
James Bergstra <bergstrj@iro.umontreal.ca>
parents: 600
diff changeset
18 return filetensor.read(f)
600
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
19
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
20 def train_10class():
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
21 #TODO: make path an option,
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
22 #TODO: make default path relative to data_root()
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
23 f = open('flickr_10classes_train.ft')
601
fd95ff96dd47 updated flickr to row-major files
James Bergstra <bergstrj@iro.umontreal.ca>
parents: 600
diff changeset
24 return filetensor.read(f)
600
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
25
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
26 def valid_10class():
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
27 #TODO: make path an option,
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
28 #TODO: make default path relative to data_root()
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
29 f = open('flickr_10classes_valid.ft')
601
fd95ff96dd47 updated flickr to row-major files
James Bergstra <bergstrj@iro.umontreal.ca>
parents: 600
diff changeset
30 return filetensor.read(f)
600
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
31
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
32 def basic_10class():
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
33 """Return the basic flickr image classification problem.
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
34 The images are 75x75, and there are 7500 training examples.
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
35 """
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
36 train = train_10class()
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
37 valid = valid_10class()
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
38 test = test_10class()
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
39
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
40 rval = Dataset()
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
41
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
42 rval.train = Dataset.Obj(
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
43 x=train[:, 0:-1],
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
44 y=numpy.asarray(train[:, -1], dtype='int64'))
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
45 rval.valid = Dataset.Obj(
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
46 x=valid[:, 0:-1],
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
47 y=numpy.asarray(valid[:, -1], dtype='int64'))
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
48 rval.test = Dataset.Obj(
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
49 x=test[:, 0:-1],
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
50 y=numpy.asarray(test[:, -1], dtype='int64'))
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
51
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
52 rval.n_classes = 10
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
53 rval.img_shape = (75,75)
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
54 return rval
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
55
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
56 def translations_10class():
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
57 raise NotImplementedError('TODO')
e56303df3c77 initial flickr
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
diff changeset
58