def handle(self, *args, **options):
qset = MaterialShape.objects.filter(
correct=True, substance__isnull=False,
substance__fail=False).order_by('-area')
for shape in queryset_progress_bar(qset):
create_shape_image_sample_task.delay(shape,
sample_width=256,
sample_height=256)
def handle(self, *args, **options):
qset = MaterialShape.objects.filter(
correct=True,
substance__isnull=False,
substance__fail=False).order_by('-area')
for shape in queryset_progress_bar(qset):
create_shape_image_sample_task.delay(
shape, sample_width=256, sample_height=256)
def handle(self, *args, **options):
larger_dimension = 320
local_dir = 'image-labels'
COLORS = [
(236, 146, 155), (127, 0, 134), (153, 224, 255), (185, 244, 105),
(157, 67, 17), (252, 45, 68), (210, 54, 255), (0, 136, 160),
(181, 189, 148), (255, 204, 180), (114, 50, 64), (195, 204, 255),
(0, 255, 255), (122, 141, 0), (255, 156, 116), (161, 0, 59),
(113, 138, 242), (0, 221, 164), (75, 72, 31), (131, 116, 111),
(255, 58, 156), (0, 76, 237), (38, 119, 97), (244, 211, 0),
(255, 60, 0), (242, 200, 232), (67, 71, 78), (0, 133, 29),
(118, 74, 0), (255, 140, 250), (0, 106, 171), (0, 248, 9),
(255, 156, 0), (114, 64, 115), (0, 179, 255), (193, 201, 189),
(213, 180, 130)
]
substance_names = [
'Wood',
'Painted',
'Fabric/cloth',
'Glass',
'Metal',
'Tile',
'Carpet/rug',
'Plastic - opaque',
'Granite/marble',
'Ceramic',
'Paper/tissue',
'Food',
'Leather',
'Plastic - clear',
'Concrete',
'Laminate',
'Brick',
'Mirror',
'Cardboard',
'Wallpaper',
'Stone',
'Wicker',
'Skin',
'Hair',
'Rubber/latex',
'Fur',
'Foliage',
'Wax',
'Linoleum',
'Sponge',
'Sky',
'Water',
'Chalkboard/blackboard',
'Cork/corkboard',
'Fire',
]
if len(COLORS) < len(substance_names):
raise ValueError("too few colors")
substance_to_id = dict(
ShapeSubstance.objects.filter(
name__in=substance_names).values_list('name', 'id'))
color_map = {
substance_to_id[name]: COLORS[idx]
for idx, name in enumerate(substance_names)
}
with open('%s/Colors.txt' % local_dir, 'w') as f:
for idx, name in enumerate(substance_names):
c = COLORS[idx]
print >> f, c[0], c[1], c[2], name
print 'Names: %s' % substance_names
print 'Name to ID: %s' % substance_to_id
print 'Color map: %s' % color_map
print 'Queueing tasks for download via celery...'
qset = Photo.objects.filter(
#scene_category__name='kitchen',
#scene_category_correct=True,
aspect_ratio__gte=0.5,
aspect_ratio__lte=2.0,
).order_by('-num_vertices') #[:2000]
for photo in queryset_progress_bar(qset):
filename_base = os.path.abspath(
os.path.join(local_dir, os.path.join('images', str(photo.id))))
photo_to_label_image_task.delay(
photo_id=photo.id,
color_map=color_map,
filename=filename_base + '_GT.png',
format='PNG',
next_task=download_photo_task.s(
photo_id=photo.id,
filename=filename_base + '.png',
format='PNG',
larger_dimension=larger_dimension))
def handle(self, *args, **options):
larger_dimension = 320
local_dir = 'image-labels'
COLORS = [
(236,146,155),
(127,0,134),
(153,224,255),
(185,244,105),
(157,67,17),
(252,45,68),
(210,54,255),
(0,136,160),
(181,189,148),
(255,204,180),
(114,50,64),
(195,204,255),
(0,255,255),
(122,141,0),
(255,156,116),
(161,0,59),
(113,138,242),
(0,221,164),
(75,72,31),
(131,116,111),
(255,58,156),
(0,76,237),
(38,119,97),
(244,211,0),
(255,60,0),
(242,200,232),
(67,71,78),
(0,133,29),
(118,74,0),
(255,140,250),
(0,106,171),
(0,248,9),
(255,156,0),
(114,64,115),
(0,179,255),
(193,201,189),
(213,180,130)
]
substance_names = [
'Wood',
'Painted',
'Fabric/cloth',
'Glass',
'Metal',
'Tile',
'Carpet/rug',
'Plastic - opaque',
'Granite/marble',
'Ceramic',
'Paper/tissue',
'Food',
'Leather',
'Plastic - clear',
'Concrete',
'Laminate',
'Brick',
'Mirror',
'Cardboard',
'Wallpaper',
'Stone',
'Wicker',
'Skin',
'Hair',
'Rubber/latex',
'Fur',
'Foliage',
'Wax',
'Linoleum',
'Sponge',
'Sky',
'Water',
'Chalkboard/blackboard',
'Cork/corkboard',
'Fire',
]
if len(COLORS) < len(substance_names):
raise ValueError("too few colors")
substance_to_id = dict(ShapeSubstance.objects
.filter(name__in=substance_names)
.values_list('name', 'id'))
color_map = {substance_to_id[name]: COLORS[idx]
for idx, name in enumerate(substance_names)}
with open('%s/Colors.txt' % local_dir, 'w') as f:
for idx, name in enumerate(substance_names):
c = COLORS[idx]
print >>f, c[0], c[1], c[2], name
print 'Names: %s' % substance_names
print 'Name to ID: %s' % substance_to_id
print 'Color map: %s' % color_map
print 'Queueing tasks for download via celery...'
qset = Photo.objects.filter(
#scene_category__name='kitchen',
#scene_category_correct=True,
aspect_ratio__gte=0.5,
aspect_ratio__lte=2.0,
).order_by('-num_vertices')#[:2000]
for photo in queryset_progress_bar(qset):
filename_base = os.path.abspath(os.path.join(
local_dir, os.path.join('images', str(photo.id))))
photo_to_label_image_task.delay(
photo_id=photo.id,
color_map=color_map,
filename=filename_base + '_GT.png',
format='PNG',
next_task=download_photo_task.s(
photo_id=photo.id,
filename=filename_base + '.png',
format='PNG',
larger_dimension=larger_dimension)
)
