def write_to_file(self, fn, filter_package=None):
write_label_vector_as_testing = False
if filter_package is not None:
write_label_vector_as_testing = filter_package[1]
sys.stderr.write(
'Info: writing image table %s; label vectors as testing? %d\n' %
(fn, write_label_vector_as_testing))
(c_total, c_written, c_neg2_but_not_testing) = (0, 0, 0)
with codecs.open(fn, 'w', encoding='UTF-8') as f:
for mir_id in sorted(self.entries):
e = self.entries[mir_id]
c_total += 1
write_this_edge = (filter_package is None) or \
(e.mir_id in filter_package[0])
if not write_this_edge:
continue
c_written += 1
f.write('%d ' % e.mir_id) # 0
f.write('%d ' % e.flickr_id) # 1
f.write('%s ' % Util.qstr(e.flickr_owner)) # 2
f.write('%s ' % Util.qstr(e.flickr_title)) # 3
f.write('%s ' % Util.qstr(e.flickr_descr)) # 4
f.write('%s ' % e.exif_data.to_table_str()) # 5,6,7
if e.flickr_locality: # 8
f.write('%s ' % Util.qstr(e.flickr_locality))
else:
f.write('none ')
v = copy.copy(e.label_vector)
if write_label_vector_as_testing:
# set all non -1 values to -2
for i in range(0, len(v)):
if v[i] != -1:
v[i] = -2
else:
# sanity check: if this is NOT testing, there should be no -2 values
if v.count(-2):
c_neg2_but_not_testing += 1
f.write('%s\n' % ','.join(map(str, v))) # 9
if c_neg2_but_not_testing != 0:
sys.stderr.write('WARNING\nWARNING\nWARNING: %s not written as testing, but contained %d (of %d) images with -2 in the label vector\nWARNING\WARNING\n'\
% (fn, c_neg2_but_not_testing, c_written))
return (c_total, c_written)
def read_from_file( fn, id_dict_to_keep=None ):
c = 0
t = ImageTable()
with codecs.open( fn, 'r', encoding='utf-8' ) as f:
while 1:
raw_line = f.readline()
if not raw_line:
break
c += 1
fields = Util.qstr_split( raw_line.strip() )
if len(fields) != 10:
raise AssertionError( 'Image table %s:%d: found %d fields, expecting 10' % (fn, c, len(fields)))
id = int( fields[0] )
if not ( (id_dict_to_keep is None) or (id in id_dict_to_keep) ):
continue
e = ImageTableEntry()
e.mir_id = int( id )
e.flickr_id = int( fields[1] )
e.flickr_owner = fields[2] if (fields[2] != 'none') else None
e.flickr_title = fields[3] if (fields[3] != 'none') else None
e.flickr_descr = fields[4] if (fields[4] != 'none') else None
if (fields[5] == 'none') and \
(fields[6] == 'none') and \
(fields[7] == 'U'):
e.exif_data = EXIFData( e.mir_id, False, 'none', 'none', 'U' )
else:
e.exif_data = EXIFData( e.mir_id, True, fields[5], fields[6], fields[7])
e.flickr_locality = fields[8] if fields[8] != 'none' else None
e.label_vector = map(int, fields[9].split(','))
t.add_entry( e )
return t
def read_from_file(fn, id_dict_to_keep=None):
c = 0
t = ImageTable()
with codecs.open(fn, 'r', encoding='utf-8') as f:
while 1:
raw_line = f.readline()
if not raw_line:
break
c += 1
fields = Util.qstr_split(raw_line.strip())
if len(fields) != 10:
raise AssertionError(
'Image table %s:%d: found %d fields, expecting 10' %
(fn, c, len(fields)))
id = int(fields[0])
if not ((id_dict_to_keep is None) or (id in id_dict_to_keep)):
continue
e = ImageTableEntry()
e.mir_id = int(id)
e.flickr_id = int(fields[1])
e.flickr_owner = fields[2] if (fields[2] != 'none') else None
e.flickr_title = fields[3] if (fields[3] != 'none') else None
e.flickr_descr = fields[4] if (fields[4] != 'none') else None
if (fields[5] == 'none') and \
(fields[6] == 'none') and \
(fields[7] == 'U'):
e.exif_data = EXIFData(e.mir_id, False, 'none', 'none',
'U')
else:
e.exif_data = EXIFData(e.mir_id, True, fields[5],
fields[6], fields[7])
e.flickr_locality = fields[8] if fields[8] != 'none' else None
e.label_vector = map(int, fields[9].split(','))
t.add_entry(e)
return t
def write_to_file( self, fn, filter_package = None ):
write_label_vector_as_testing = False
if filter_package is not None:
write_label_vector_as_testing = filter_package[1]
sys.stderr.write('Info: writing image table %s; label vectors as testing? %d\n' % (fn, write_label_vector_as_testing ))
(c_total, c_written, c_neg2_but_not_testing) = (0,0,0)
with codecs.open( fn, 'w', encoding='UTF-8' ) as f:
for mir_id in sorted( self.entries ):
e = self.entries[ mir_id ]
c_total += 1
write_this_edge = (filter_package is None) or \
(e.mir_id in filter_package[0])
if not write_this_edge:
continue
c_written += 1
f.write( '%d ' % e.mir_id ) # 0
f.write( '%d ' % e.flickr_id ) # 1
f.write( '%s ' % Util.qstr( e.flickr_owner )) # 2
f.write( '%s ' % Util.qstr( e.flickr_title )) # 3
f.write( '%s ' % Util.qstr( e.flickr_descr )) # 4
f.write( '%s ' % e.exif_data.to_table_str() ) # 5,6,7
if e.flickr_locality: # 8
f.write( '%s ' % Util.qstr( e.flickr_locality ))
else:
f.write( 'none ' )
v = copy.copy( e.label_vector )
if write_label_vector_as_testing:
# set all non -1 values to -2
for i in range(0, len(v)):
if v[i] != -1:
v[i] = -2
else:
# sanity check: if this is NOT testing, there should be no -2 values
if v.count(-2):
c_neg2_but_not_testing += 1
f.write( '%s\n' % ','.join(map(str, v))) # 9
if c_neg2_but_not_testing != 0:
sys.stderr.write('WARNING\nWARNING\nWARNING: %s not written as testing, but contained %d (of %d) images with -2 in the label vector\nWARNING\WARNING\n'\
% (fn, c_neg2_but_not_testing, c_written))
return (c_total, c_written)
def write_to_file( self, fn ):
with open( fn, 'w' ) as f:
f.write( '%d %d\n' % (len(self.group_text_lut), len(self.tag_word_text_lut)))
for i in sorted( self.group_text_lut.iteritems(), key=lambda x:x[1] ):
(entry_id, entry_text) = (i[1], i[0])
f.write( '%d G %s\n' % ( entry_id, Util.qstr( entry_text )))
for i in sorted( self.tag_word_text_lut.iteritems(), key=lambda x:x[1] ):
(entry_id, entry_text) = (i[1], i[0])
f.write( '%d %s %s\n' % ( entry_id, self.tag_word_text_src[ entry_id], Util.qstr( entry_text )))
def read_from_file(fn):
L2I = dict()
I2L = dict()
with open(fn, 'r') as f:
while 1:
raw_line = f.readline()
if not raw_line:
break
fields = Util.qstr_split(raw_line.strip())
I2L[int(fields[0])] = fields[1]
L2I[fields[1]] = int(fields[0])
return LabelTable(L2I, I2L)
def read_from_file( fn ):
t = ImageIndicatorLookupTable()
with open( fn, 'r' ) as f:
header_fields = Util.qstr_split( f.readline().strip() )
if len(header_fields) != 2:
raise AssertionError( 'ImageIndicatorLookupTable "%s": header had %d fields, expected 2' % \
( fn, len(header_fields)))
(n_groups, n_words_and_tags) = map(int, header_fields)
for i in range(0, n_groups):
fields = Util.qstr_split( f.readline().strip() )
if len(fields) != 3:
raise AssertionError( 'ImageIndicatorLookupTable "%s": group %d had %d fields, expected 3' % \
(fn, i, len(fields)))
if fields[1] != 'G':
raise AssertionError( 'ImageIndicatorLookupTable "%s": entry %d flavor was %s; expected "G"' % \
(fn, i, fields[1]))
t.group_text_lut[ fields[2] ] = int( fields[0] )
for i in range(0, n_words_and_tags):
fields = Util.qstr_split( f.readline().strip() )
if len(fields) != 3:
raise AssertionError( 'ImageIndicatorLookupTable "%s": word %d had %d fields, expected 3' % \
(fn, i, len(word_fields)))
if (fields[1] != 'W') and (fields[1] != 'T') and (fields[1] != 'B'):
raise AssertionError( 'ImageIndicatorLookupTable "%s": entry %d flavor was %s; expected "W", "T", or "B"' % \
(fn, i, fields[1]))
t.tag_word_text_lut[ fields[2] ] = int( fields[0] )
t.tag_word_text_src[ int(fields[0] ) ] = fields[1]
sys.stderr.write('Info: IILUT read %d / %d / %d / %d groups / tags / words / both\n' % \
(len(t.group_text_lut), t.tag_word_text_src.values().count('T'), \
t.tag_word_text_src.values().count('W'), \
t.tag_word_text_src.values().count('B')))
return t
g = Graph()
g.add_vertices( len(nodes))
graph_edges = list( [node_id_map[e.image_A_id], node_id_map[e.image_B_id]] \
for e in edges if edges_predicate(e))
g.add_edges( graph_edges)
layout = g.layout_circle()
# adding text to plot courtesy http://stackoverflow.com/questions/18250684/add-title-and-legend-to-igraph-plots
plot = Plot(output_fn, bbox=[2000,2000])
plot.add( g, layout=layout, bbox=[50,50,1950,1950])
ctx = cairo.Context( plot.surface )
ctx.set_font_size(36)
drawer = TextDrawer( ctx, "Label: '%s'\n%s\n%d nodes, %d edges" % \
(LABEL, tag, len(nodes), len(graph_edges)), halign=TextDrawer.RIGHT)
drawer.draw_at(10,1900,width=1900)
plot.save()
sys.stderr.write("Info: wrote %s\n" % output_fn)
p = Paths()
nodes = Util.nodes_with_label(p, PHASE, LABEL)
edges = CP6ImageEdge.edges_in_nodeset( p, PHASE, nodes)
# map node and edge IDs to graph vertex ordinals
node_id_map = dict( [val, idx] for (idx, val) in enumerate(nodes))
make_graph( node_id_map, nodes, edges, shared_groups_only, "%s.group-only.png" % LABEL, "shared groups only")
make_graph( node_id_map, nodes, edges, shared_flags_only, "%s.flags-only.png" % LABEL, "same user/location/contact only")
make_graph( node_id_map, nodes, edges, all_edges, "%s.all-edges.png" % LABEL, "any relationship")
for n in (1,2,5,10,15,20):
make_graph( node_id_map, nodes, edges, functools.partial(only_N_shared_words, N=n), "%s.%02d-shared-word.png" % (LABEL, n) , "Exactly %d shared word(s)" % n)
def write_to_file(self, fn):
with open(fn, 'w') as f:
for index in self.idset:
f.write('%d %s\n' % (index, Util.qstr(self.id2label[index])))