class TinyImage:
def __init__(self):
self.config = config.config()
self.meta = BlockStore(self.config['tinyimages']['metapath'], 768)
self.data = BlockStore(self.config['tinyimages']['datapath'], 3072)
self.img_count = 79302017
#public functions
def byid(self, ids):
if isinstance(ids, int):
return numpy.fromstring(self.data.byid(ids), dtype='uint8')
elif isinstance(ids, tuple):
o = []
for s in self.data.slice(ids[0], ids[1]):
o.append(numpy.fromstring(s, dtype='uint8'))
return o
else:
o = []
for i in ids:
o.append(numpy.fromstring(self.data.byid(i), dtype='uint8'))
return o
def display(self, items):
import cStringIO as StringIO
import base64
from IPython.core.display import HTML
output_html = ""
for i in items:
t = i.reshape(32,32,3, order="F").copy()
img = scipy.misc.toimage(t)
output = StringIO.StringIO()
img.save(output, format="PNG")
output_html += '<img src="data:image/png;base64,%s"/>' % base64.b64encode(output.getvalue())
return HTML(output_html)
def search(self, keyword, limit):
(l, h) = self._logSearch(keyword)
found = False
found_count = 0
o = []
for i in range(l, h):
curr_word = self._keywordFromMeta(i)
if curr_word.lower() == keyword.lower():
found = True
o.append(i)
found_count += 1
if (found_count == limit):
break
else:
if (found):
break
return o
def _keywordFromMeta(self, index):
for s in self.meta.slice(index, index):
return s[0:80].strip()
def _logSearch(self, term):
low = 0
high = self.img_count
for i in range(0, 9):
curr_word = self._keywordFromMeta(int((low + high) / 2))
cmp = _strcmp(curr_word.lower(), term.lower())
if (cmp == 0):
return (low, high)
if (cmp == 1):
high = ((low + high) / 2)
if (cmp == -1):
low = ((low + high) / 2)
return (low, high)
def subsets(self):
return None
class TinyImage:
def __init__(self):
self.config = config.config()
self.meta = BlockStore(self.config['tinyimages']['metapath'], 768)
self.data = BlockStore(self.config['tinyimages']['datapath'], 3072)
self.img_count = 79302017
#public functions
def byid(self, ids):
if isinstance(ids, int):
return numpy.fromstring(self.data.byid(ids), dtype='uint8')
elif isinstance(ids, tuple):
o = []
for s in self.data.slice(ids[0], ids[1]):
o.append(numpy.fromstring(s, dtype='uint8'))
return o
else:
o = []
for i in ids:
o.append(numpy.fromstring(self.data.byid(i), dtype='uint8'))
return o
def display(self, items):
import cStringIO as StringIO
import base64
from IPython.core.display import HTML
output_html = ""
for i in items:
t = i.reshape(32, 32, 3, order="F").copy()
img = scipy.misc.toimage(t)
output = StringIO.StringIO()
img.save(output, format="PNG")
output_html += '<img src="data:image/png;base64,%s"/>' % base64.b64encode(
output.getvalue())
return HTML(output_html)
def search(self, keyword, limit):
(l, h) = self._logSearch(keyword)
found = False
found_count = 0
o = []
for i in range(l, h):
curr_word = self._keywordFromMeta(i)
if curr_word.lower() == keyword.lower():
found = True
o.append(i)
found_count += 1
if (found_count == limit):
break
else:
if (found):
break
return o
def _keywordFromMeta(self, index):
for s in self.meta.slice(index, index):
return s[0:80].strip()
def _logSearch(self, term):
low = 0
high = self.img_count
for i in range(0, 9):
curr_word = self._keywordFromMeta(int((low + high) / 2))
cmp = _strcmp(curr_word.lower(), term.lower())
if (cmp == 0):
return (low, high)
if (cmp == 1):
high = ((low + high) / 2)
if (cmp == -1):
low = ((low + high) / 2)
return (low, high)
def subsets(self):
return None
