Add default colors for masks (#1549)

* Use default colors for masks

* do not change labels in voc converter

* update tests

* add backgound color handling

* consider case with existing source colormap

* update changelog

* fix file name

* fix insert

* move color table colors to beginning

CHANGELOG.md

@@ -7,6 +7,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 ## [1.1.0-alpha] - Unreleased
 ### Added
 - Throttling policy for unauthenticated users (<https://github.com/opencv/cvat/pull/1531>)
+- Added default label color table for mask export (https://github.com/opencv/cvat/pull/1549)
 
 ### Changed
 - Removed information about e-mail from the basic user information (<https://github.com/opencv/cvat/pull/1627>)

cvat/apps/dataset_manager/formats/mask.py

@@ -2,6 +2,7 @@
 #
 # SPDX-License-Identifier: MIT
 
+import os.path as osp
 from tempfile import TemporaryDirectory
 
 from pyunpack import Archive
@@ -9,7 +10,9 @@ from pyunpack import Archive
 from cvat.apps.dataset_manager.bindings import (CvatTaskDataExtractor,
     import_dm_annotations)
 from cvat.apps.dataset_manager.util import make_zip_archive
+from datumaro.cli.util import make_file_name
 from datumaro.components.project import Dataset
+from datumaro.util.mask_tools import generate_colormap
 
 from .registry import dm_env, exporter, importer
 
@@ -25,7 +28,8 @@ def _export(dst_file, task_data, save_images=False):
     extractor = Dataset.from_extractors(extractor) # apply lazy transforms
     with TemporaryDirectory() as temp_dir:
         converter = dm_env.make_converter('voc_segmentation',
-            apply_colormap=True, label_map='source', save_images=save_images)
+            apply_colormap=True, label_map=make_colormap(task_data),
+            save_images=save_images)
         converter(extractor, save_dir=temp_dir)
 
         make_zip_archive(temp_dir, dst_file)
@@ -39,3 +43,43 @@ def _import(src_file, task_data):
         masks_to_polygons = dm_env.transforms.get('masks_to_polygons')
         dataset = dataset.transform(masks_to_polygons)
         import_dm_annotations(dataset, task_data)
+
+
+DEFAULT_COLORMAP_CAPACITY = 2000
+DEFAULT_COLORMAP_PATH = osp.join(osp.dirname(__file__), 'predefined_colors.txt')
+def parse_default_colors(file_path=None):
+    if file_path is None:
+        file_path = DEFAULT_COLORMAP_PATH
+
+    colors = {}
+    with open(file_path) as f:
+        for line in f:
+            line = line.strip()
+            if not line or line[0] == '#':
+                continue
+            _, label, color = line.split(':')
+            colors[label] = tuple(map(int, color.split(',')))
+    return colors
+
+def normalize_label(label):
+    label = make_file_name(label) # basically, convert to ASCII lowercase
+    label = label.replace('-', '_')
+    return label
+
+def make_colormap(task_data):
+    labels = sorted([label['name']
+        for _, label in task_data.meta['task']['labels']])
+    if 'background' not in labels:
+        labels.insert(0, 'background')
+
+    predefined = parse_default_colors()
+
+    # NOTE: using pop() to avoid collisions
+    colormap = {k: predefined.pop(normalize_label(k), None) for k in labels}
+
+    random_labels = [k for k in labels if not colormap[k]]
+    if random_labels:
+        colors = generate_colormap(DEFAULT_COLORMAP_CAPACITY + len(random_labels))
+        for i, label in enumerate(random_labels):
+            colormap[label] = colors[DEFAULT_COLORMAP_CAPACITY + i]
+    return {l: [c, [], []] for l, c in colormap.items()}

cvat/apps/dataset_manager/formats/predefined_colors.txt

@@ -0,0 +1,731 @@
+# ImageNet + OpenImages + PASCAL VOC +
+# CamVid + CityScapes + Kitti +
+# COCO + COCO stuff
+
+# Format:
+# pascal idx : normalized label name : R,G,B
+
+0:background:0,0,0
+1:accordion:128,0,0
+2:adhesive_tape:0,128,0
+3:aeroplane:128,128,0
+4:aircraft:0,0,128
+5:airplane:128,0,128
+6:alarm_clock:0,128,128
+7:alpaca:128,128,128
+8:ambulance:64,0,0
+9:animal:192,0,0
+10:ant:64,128,0
+11:antelope:192,128,0
+12:apple:64,0,128
+13:armadillo:192,0,128
+14:artichoke:64,128,128
+15:asparagus:192,128,128
+16:axe:0,64,0
+17:baby_bed:128,64,0
+18:backpack:0,192,0
+19:bagel:128,192,0
+20:balance_beam:0,64,128
+21:ball:128,64,128
+22:balloon:0,192,128
+23:banana:128,192,128
+24:band_aid:64,64,0
+25:banjo:192,64,0
+26:banner:64,192,0
+27:barge:192,192,0
+28:barrel:64,64,128
+29:baseball:192,64,128
+30:baseball_bat:64,192,128
+31:baseball_glove:192,192,128
+32:basketball:0,0,64
+33:bat:128,0,64
+34:bathing_cap:0,128,64
+35:bathroom_cabinet:128,128,64
+36:bathtub:0,0,192
+37:beaker:128,0,192
+38:bear:0,128,192
+39:bed:128,128,192
+40:bee:64,0,64
+41:beehive:192,0,64
+42:beer:64,128,64
+43:beetle:192,128,64
+44:bell_pepper:64,0,192
+45:belt:192,0,192
+46:bench:64,128,192
+47:bg:192,128,192
+48:bicycle:0,64,64
+49:bicycle_helmet:128,64,64
+50:bicycle_wheel:0,192,64
+51:bidet:128,192,64
+52:billboard:0,64,192
+53:billiard_table:128,64,192
+54:binder:0,192,192
+55:binoculars:128,192,192
+56:bird:64,64,64
+57:blanket:192,64,64
+58:blender:64,192,64
+59:blue_jay:192,192,64
+60:boat:64,64,192
+61:book:192,64,192
+62:bookcase:64,192,192
+63:bookshelf:192,192,192
+64:boot:32,0,0
+65:bottle:160,0,0
+66:bow:32,128,0
+67:bow_and_arrow:160,128,0
+68:bow_tie:32,0,128
+69:bowl:160,0,128
+70:box:32,128,128
+71:boy:160,128,128
+72:branch:96,0,0
+73:brassiere:224,0,0
+74:bread:96,128,0
+75:bridge:224,128,0
+76:briefcase:96,0,128
+77:broccoli:224,0,128
+78:bronze_sculpture:96,128,128
+79:brown_bear:224,128,128
+80:building:32,64,0
+81:building_other:160,64,0
+82:bull:32,192,0
+83:burrito:160,192,0
+84:bus:32,64,128
+85:bush:160,64,128
+86:bust:32,192,128
+87:butterfly:160,192,128
+88:cabbage:96,64,0
+89:cabinet:224,64,0
+90:cabinetry:96,192,0
+91:cage:224,192,0
+92:cake:96,64,128
+93:cake_stand:224,64,128
+94:camel:96,192,128
+95:camera:224,192,128
+96:can_opener:32,0,64
+97:canary:160,0,64
+98:candle:32,128,64
+99:candy:160,128,64
+100:cannon:32,0,192
+101:canoe:160,0,192
+102:car:32,128,192
+103:caravan:160,128,192
+104:cardboard:96,0,64
+105:carnivore:224,0,64
+106:carpet:96,128,64
+107:carrot:224,128,64
+108:cart:96,0,192
+109:castle:224,0,192
+110:cat:96,128,192
+111:caterpillar:224,128,192
+112:cattle:32,64,64
+113:ceiling_fan:160,64,64
+114:ceiling_other:32,192,64
+115:ceiling_tile:160,192,64
+116:cell_phone:32,64,192
+117:cello:160,64,192
+118:centipede:32,192,192
+119:chain_saw:160,192,192
+120:chair:96,64,64
+121:cheetah:224,64,64
+122:chest_of_drawers:96,192,64
+123:chicken:224,192,64
+124:chime:96,64,192
+125:chopsticks:224,64,192
+126:christmas_tree:96,192,192
+127:clock:224,192,192
+128:cloth:0,32,0
+129:clothes:128,32,0
+130:clouds:0,160,0
+131:coat:128,160,0
+132:cocktail:0,32,128
+133:cocktail_shaker:128,32,128
+134:coconut:0,160,128
+135:coffee:128,160,128
+136:coffee_cup:64,32,0
+137:coffee_maker:192,32,0
+138:coffee_table:64,160,0
+139:coffeemaker:192,160,0
+140:coin:64,32,128
+141:common_fig:192,32,128
+142:computer_keyboard:64,160,128
+143:computer_monitor:192,160,128
+144:computer_mouse:0,96,0
+145:construction:128,96,0
+146:convenience_store:0,224,0
+147:cookie:128,224,0
+148:corded_phone:0,96,128
+149:corkscrew:128,96,128
+150:couch:0,224,128
+151:counter:128,224,128
+152:countertop:64,96,0
+153:cow:192,96,0
+154:cowboy_hat:64,224,0
+155:crab:192,224,0
+156:cream:64,96,128
+157:cricket_ball:192,96,128
+158:crocodile:64,224,128
+159:croissant:192,224,128
+160:croquet_ball:0,32,64
+161:crosswalk:128,32,64
+162:crown:0,160,64
+163:crutch:128,160,64
+164:cucumber:0,32,192
+165:cup:128,32,192
+166:cup_or_mug:0,160,192
+167:cupboard:128,160,192
+168:curtain:64,32,64
+169:cutting_board:192,32,64
+170:cyclist:64,160,64
+171:dagger:192,160,64
+172:deer:64,32,192
+173:desk:192,32,192
+174:desk_stuff:64,160,192
+175:dessert:192,160,192
+176:diaper:0,96,64
+177:dice:128,96,64
+178:digital_clock:0,224,64
+179:dining_table:128,224,64
+180:diningtable:0,96,192
+181:dinosaur:128,96,192
+182:dirt:0,224,192
+183:dishwasher:128,224,192
+184:dog:64,96,64
+185:dog_bed:192,96,64
+186:doll:64,224,64
+187:dolphin:192,224,64
+188:domestic_cat:64,96,192
+189:dontcare:192,96,192
+190:donut:64,224,192
+191:door:192,224,192
+192:door_handle:32,32,0
+193:door_stuff:160,32,0
+194:doughnut:32,160,0
+195:dragonfly:160,160,0
+196:drawer:32,32,128
+197:dress:160,32,128
+198:drink:32,160,128
+199:drinking_straw:160,160,128
+200:drum:96,32,0
+201:duck:224,32,0
+202:dumbbell:96,160,0
+203:dynamic:224,160,0
+204:eagle:96,32,128
+205:earrings:224,32,128
+206:egg:96,160,128
+207:electric_fan:224,160,128
+208:elephant:32,96,0
+209:envelope:160,96,0
+210:eye_glasses:32,224,0
+211:face_powder:160,224,0
+212:falcon:32,96,128
+213:fedora:160,96,128
+214:fence:32,224,128
+215:fence_guard:160,224,128
+216:fig:96,96,0
+217:filing_cabinet:224,96,0
+218:finish:96,224,0
+219:fire_hydrant:224,224,0
+220:fireplace:96,96,128
+221:fish:224,96,128
+222:flag:96,224,128
+223:flashlight:224,224,128
+224:floor_marble:32,32,64
+225:floor_other:160,32,64
+226:floor_stone:32,160,64
+227:floor_tile:160,160,64
+228:floor_wood:32,32,192
+229:flower:160,32,192
+230:flower_pot:32,160,192
+231:flowerpot:160,160,192
+232:flute:96,32,64
+233:fog:224,32,64
+234:food_other:96,160,64
+235:food_processor:224,160,64
+236:football:96,32,192
+237:football_helmet:224,32,192
+238:footwear:96,160,192
+239:fork:224,160,192
+240:fountain:32,96,64
+241:fox:160,96,64
+242:french_fries:32,224,64
+243:french_horn:160,224,64
+244:frisbee:32,96,192
+245:frog:160,96,192
+246:fruit:32,224,192
+247:frying_pan:160,224,192
+248:furniture:96,96,64
+249:furniture_other:224,96,64
+250:gas_stove:96,224,64
+251:giant_panda:224,224,64
+252:giraffe:96,96,192
+253:girl:224,96,192
+254:glasses:96,224,192
+255:glove:224,224,192
+256:goat:0,0,32
+257:goggles:128,0,32
+258:goldfish:0,128,32
+259:golf_ball:128,128,32
+260:golf_cart:0,0,160
+261:golfcart:128,0,160
+262:gondola:0,128,160
+263:goose:128,128,160
+264:grape:64,0,32
+265:grapefruit:192,0,32
+266:grass:64,128,32
+267:gravel:192,128,32
+268:ground:64,0,160
+269:ground_other:192,0,160
+270:guacamole:64,128,160
+271:guitar:192,128,160
+272:hair_brush:0,64,32
+273:hair_drier:128,64,32
+274:hair_dryer:0,192,32
+275:hair_spray:128,192,32
+276:hamburger:0,64,160
+277:hammer:128,64,160
+278:hamster:0,192,160
+279:handbag:128,192,160
+280:handgun:64,64,32
+281:harbor_seal:192,64,32
+282:harmonica:64,192,32
+283:harp:192,192,32
+284:harpsichord:64,64,160
+285:hat:192,64,160
+286:hat_with_a_wide_brim:64,192,160
+287:head_cabbage:192,192,160
+288:headphones:0,0,96
+289:helicopter:128,0,96
+290:helmet:0,128,96
+291:high_heels:128,128,96
+292:hill:0,0,224
+293:hippopotamus:128,0,224
+294:home_appliance:0,128,224
+295:honeycomb:128,128,224
+296:horizontal_bar:64,0,96
+297:horn:192,0,96
+298:horse:64,128,96
+299:hot_dog:192,128,96
+300:hotdog:64,0,224
+301:house:192,0,224
+302:houseplant:64,128,224
+303:human:192,128,224
+304:human_arm:0,64,96
+305:human_beard:128,64,96
+306:human_ear:0,192,96
+307:human_eye:128,192,96
+308:human_face:0,64,224
+309:human_foot:128,64,224
+310:human_hair:0,192,224
+311:human_hand:128,192,224
+312:human_head:64,64,96
+313:human_leg:192,64,96
+314:human_mouth:64,192,96
+315:human_nose:192,192,96
+316:ice_cream:64,64,224
+317:ignore:192,64,224
+318:infant_bed:64,192,224
+319:insect:192,192,224
+320:invertebrate:32,0,32
+321:ipod:160,0,32
+322:isopod:32,128,32
+323:jacket:160,128,32
+324:jaguar:32,0,160
+325:jeans:160,0,160
+326:jellyfish:32,128,160
+327:jet_ski:160,128,160
+328:jug:96,0,32
+329:juice:224,0,32
+330:kangaroo:96,128,32
+331:kettle:224,128,32
+332:keyboard:96,0,160
+333:kitchen_appliance:224,0,160
+334:kitchen_dining_room_table:96,128,160
+335:kitchen_knife:224,128,160
+336:kite:32,64,32
+337:knife:160,64,32
+338:koala_bear:32,192,32
+339:ladder:160,192,32
+340:ladle:32,64,160
+341:ladybug:160,64,160
+342:lamp:32,192,160
+343:land_vehicle:160,192,160
+344:lantern:96,64,32
+345:laptop:224,64,32
+346:lavender:96,192,32
+347:leaves:224,192,32
+348:lemon:96,64,160
+349:leopard:224,64,160
+350:license_plate:96,192,160
+351:lifejacket:224,192,160
+352:light:32,0,96
+353:light_bulb:160,0,96
+354:light_switch:32,128,96
+355:lighthouse:160,128,96
+356:lily:32,0,224
+357:limousine:160,0,224
+358:lion:32,128,224
+359:lipstick:160,128,224
+360:lizard:96,0,96
+361:lobster:224,0,96
+362:loveseat:96,128,96
+363:luggage_and_bags:224,128,96
+364:lynx:96,0,224
+365:maillot:224,0,224
+366:man:96,128,224
+367:mango:224,128,224
+368:maple:32,64,96
+369:maraca:160,64,96
+370:marine_invertebrates:32,192,96
+371:marine_mammal:160,192,96
+372:mat:32,64,224
+373:measuring_cup:160,64,224
+374:mechanical_fan:32,192,224
+375:metal:160,192,224
+376:microphone:96,64,96
+377:microwave:224,64,96
+378:microwave_oven:96,192,96
+379:milk_can:224,192,96
+380:miniskirt:96,64,224
+381:mirror:224,64,224
+382:mirror_stuff:96,192,224
+383:misc:224,192,224
+384:missile:0,32,32
+385:mixer:128,32,32
+386:mobile_phone:0,160,32
+387:monkey:128,160,32
+388:moss:0,32,160
+389:moths_and_butterflies:128,32,160
+390:motorbike:0,160,160
+391:motorcycle:128,160,160
+392:mountain:64,32,32
+393:mouse:192,32,32
+394:mud:64,160,32
+395:muffin:192,160,32
+396:mug:64,32,160
+397:mule:192,32,160
+398:mushroom:64,160,160
+399:musical_instrument:192,160,160
+400:musical_keyboard:0,96,32
+401:nail:128,96,32
+402:napkin:0,224,32
+403:nature:128,224,32
+404:neck_brace:0,96,160
+405:necklace:128,96,160
+406:net:0,224,160
+407:nightstand:128,224,160
+408:non_vehicle:64,96,32
+409:obj:192,96,32
+410:object:64,224,32
+411:oboe:192,224,32
+412:office_building:64,96,160
+413:office_supplies:192,96,160
+414:on_rails:64,224,160
+415:orange:192,224,160
+416:organ:0,32,96
+417:ostrich:128,32,96
+418:otter:0,160,96
+419:oven:128,160,96
+420:owl:0,32,224
+421:oyster:128,32,224
+422:paddle:0,160,224
+423:palm_tree:128,160,224
+424:pancake:64,32,96
+425:paper:192,32,96
+426:paper_towel:64,160,96
+427:parachute:192,160,96
+428:parking:64,32,224
+429:parking_meter:192,32,224
+430:parrot:64,160,224
+431:pasta:192,160,224
+432:pavement:0,96,96
+433:peach:128,96,96
+434:pear:0,224,96
+435:pedestrian:128,224,96
+436:pen:0,96,224
+437:pencil_box:128,96,224
+438:pencil_sharpener:0,224,224
+439:penguin:128,224,224
+440:perfume:64,96,96
+441:person:192,96,96
+442:personal_care:64,224,96
+443:piano:192,224,96
+444:picnic_basket:64,96,224
+445:picture_frame:192,96,224
+446:pig:64,224,224
+447:pillow:192,224,224
+448:pineapple:32,32,32
+449:ping_pong_ball:160,32,32
+450:pitcher:32,160,32
+451:pizza:160,160,32
+452:plant_other:32,32,160
+453:plastic:160,32,160
+454:plastic_bag:32,160,160
+455:plate:160,160,160
+456:plate_rack:96,32,32
+457:platform:224,32,32
+458:platter:96,160,32
+459:playingfield:224,160,32
+460:plumbing_fixture:96,32,160
+461:polar_bear:224,32,160
+462:pole:96,160,160
+463:pole_group:224,160,160
+464:pomegranate:32,96,32
+465:popcorn:160,96,32
+466:popsicle:32,224,32
+467:porch:160,224,32
+468:porcupine:32,96,160
+469:poster:160,96,160
+470:potato:32,224,160
+471:potted_plant:160,224,160
+472:pottedplant:96,96,32
+473:power_drill:224,96,32
+474:power_plugs_and_sockets:96,224,32
+475:pressure_cooker:224,224,32
+476:pretzel:96,96,160
+477:printer:224,96,160
+478:puck:96,224,160
+479:pumpkin:224,224,160
+480:punching_bag:32,32,96
+481:purse:160,32,96
+482:rabbit:32,160,96
+483:raccoon:160,160,96
+484:racket:32,32,224
+485:radish:160,32,224
+486:rail:32,160,224
+487:rail_track:160,160,224
+488:railing:96,32,96
+489:railroad:224,32,96
+490:raven:96,160,96
+491:ray:224,160,96
+492:red_panda:96,32,224
+493:refrigerator:224,32,224
+494:remote:96,160,224
+495:remote_control:224,160,224
+496:reptile:32,96,96
+497:rhinoceros:160,96,96
+498:rider:32,224,96
+499:rifle:160,224,96
+500:ring_binder:32,96,224
+501:river:160,96,224
+502:road:32,224,224
+503:rock:160,224,224
+504:rocket:96,96,96
+505:roller_skates:224,96,96
+506:roof:96,224,96
+507:rose:224,224,96
+508:rubber_eraser:96,96,224
+509:rug:224,96,224
+510:rugby_ball:96,224,224
+511:ruler:224,224,224
+512:salad:16,0,0
+513:salt_and_pepper_shakers:144,0,0
+514:salt_or_pepper_shaker:16,128,0
+515:sand:144,128,0
+516:sandal:16,0,128
+517:sandwich:144,0,128
+518:saucer:16,128,128
+519:saxophone:144,128,128
+520:scarf:80,0,0
+521:scissors:208,0,0
+522:scoreboard:80,128,0
+523:scorpion:208,128,0
+524:screwdriver:80,0,128
+525:sculpture:208,0,128
+526:sea:80,128,128
+527:sea_lion:208,128,128
+528:sea_turtle:16,64,0
+529:seafood:144,64,0
+530:seahorse:16,192,0
+531:seal:144,192,0
+532:seat_belt:16,64,128
+533:segway:144,64,128
+534:serving_tray:16,192,128
+535:sewing_machine:144,192,128
+536:shark:80,64,0
+537:sheep:208,64,0
+538:shelf:80,192,0
+539:shellfish:208,192,0
+540:shirt:80,64,128
+541:shoe:208,64,128
+542:shorts:80,192,128
+543:shotgun:208,192,128
+544:shower:16,0,64
+545:shrimp:144,0,64
+546:sidewalk:16,128,64
+547:sink:144,128,64
+548:skateboard:16,0,192
+549:ski:144,0,192
+550:skirt:16,128,192
+551:skis:144,128,192
+552:skull:80,0,64
+553:skunk:208,0,64
+554:sky:80,128,64
+555:sky_other:208,128,64
+556:skyscraper:80,0,192
+557:slow_cooker:208,0,192
+558:snail:80,128,192
+559:snake:208,128,192
+560:snow:16,64,64
+561:snowboard:144,64,64
+562:snowman:16,192,64
+563:snowmobile:144,192,64
+564:snowplow:16,64,192
+565:soap_dispenser:144,64,192
+566:soccer_ball:16,192,192
+567:sock:144,192,192
+568:sofa:80,64,64
+569:sofa_bed:208,64,64
+570:solid_other:80,192,64
+571:sombrero:208,192,64
+572:sparrow:80,64,192
+573:spatula:208,64,192
+574:spider:80,192,192
+575:spoon:208,192,192
+576:sports_ball:48,0,0
+577:sports_uniform:176,0,0
+578:squash:48,128,0
+579:squirrel:176,128,0
+580:stairs:48,0,128
+581:starfish:176,0,128
+582:start:48,128,128
+583:static:176,128,128
+584:stationary_bicycle:112,0,0
+585:stethoscope:240,0,0
+586:stone:112,128,0
+587:stool:240,128,0
+588:stop:112,0,128
+589:stop_sign:240,0,128
+590:stove:112,128,128
+591:strainer:240,128,128
+592:straw:48,64,0
+593:strawberry:176,64,0
+594:street_light:48,192,0
+595:street_sign:176,192,0
+596:stretcher:48,64,128
+597:structural_other:176,64,128
+598:studio_couch:48,192,128
+599:submarine_sandwich:176,192,128
+600:suit:112,64,0
+601:suitcase:240,64,0
+602:sun_hat:112,192,0
+603:sunflower:240,192,0
+604:sunglasses:112,64,128
+605:surfboard:240,64,128
+606:sushi:112,192,128
+607:swan:240,192,128
+608:swim_cap:48,0,64
+609:swimming_pool:176,0,64
+610:swimming_trunks:48,128,64
+611:swimwear:176,128,64
+612:swine:48,0,192
+613:sword:176,0,192
+614:syringe:48,128,192
+615:table:176,128,192
+616:table_tennis_racket:112,0,64
+617:tablet_computer:240,0,64
+618:tableware:112,128,64
+619:taco:240,128,64
+620:tank:112,0,192
+621:tap:240,0,192
+622:tape_player:112,128,192
+623:tart:240,128,192
+624:taxi:48,64,64
+625:tea:176,64,64
+626:teapot:48,192,64
+627:teddy_bear:176,192,64
+628:telephone:48,64,192
+629:television:176,64,192
+630:tennis_ball:48,192,192
+631:tennis_racket:176,192,192
+632:tent:112,64,64
+633:terrain:240,64,64
+634:textile_other:112,192,64
+635:tiara:240,192,64
+636:tick:112,64,192
+637:tie:240,64,192
+638:tiger:112,192,192
+639:tin_can:240,192,192
+640:tire:16,32,0
+641:toaster:144,32,0
+642:toilet:16,160,0
+643:toilet_paper:144,160,0
+644:tomato:16,32,128
+645:toothbrush:144,32,128
+646:torch:16,160,128
+647:tortoise:144,160,128
+648:towel:80,32,0
+649:tower:208,32,0
+650:toy:80,160,0
+651:traffic_light:208,160,0
+652:traffic_sign:80,32,128
+653:trailer:208,32,128
+654:train:80,160,128
+655:tram:208,160,128
+656:treadmill:16,96,0
+657:tree:144,96,0
+658:tripod:16,224,0
+659:trombone:144,224,0
+660:trousers:16,96,128
+661:truck:144,96,128
+662:trumpet:16,224,128
+663:tunnel:144,224,128
+664:turkey:80,96,0
+665:turtle:208,96,0
+666:tv:80,224,0
+667:tv_or_monitor:208,224,0
+668:tvmonitor:80,96,128
+669:umbrella:208,96,128
+670:undefined:80,224,128
+671:unicycle:208,224,128
+672:unlabeled:16,32,64
+673:vacuum:144,32,64
+674:van:16,160,64
+675:vase:144,160,64
+676:vegetable:16,32,192
+677:vegetation:144,32,192
+678:vehicle:16,160,192
+679:vehicle_registration_plate:144,160,192
+680:violin:80,32,64
+681:void:208,32,64
+682:volleyball:80,160,64
+683:waffle:208,160,64
+684:waffle_iron:80,32,192
+685:wall:208,32,192
+686:wall_brick:80,160,192
+687:wall_clock:208,160,192
+688:wall_concrete:16,96,64
+689:wall_other:144,96,64
+690:wall_panel:16,224,64
+691:wall_stone:144,224,64
+692:wall_tile:16,96,192
+693:wall_wood:144,96,192
+694:washer:16,224,192
+695:washing_machine:144,224,192
+696:waste_container:80,96,64
+697:watch:208,96,64
+698:water_bottle:80,224,64
+699:water_other:208,224,64
+700:watercraft:80,96,192
+701:waterdrops:208,96,192
+702:watermelon:80,224,192
+703:weapon:208,224,192
+704:whale:48,32,0
+705:wheel:176,32,0
+706:wheelchair:48,160,0
+707:whiteboard:176,160,0
+708:willow:48,32,128
+709:window:176,32,128
+710:window_blind:48,160,128
+711:window_other:176,160,128
+712:wine:112,32,0
+713:wine_bottle:240,32,0
+714:wine_glass:112,160,0
+715:winter_melon:240,160,0
+716:wok:112,32,128
+717:woman:240,32,128
+718:wood:112,160,128
+719:wood_burning_stove:240,160,128
+720:woodpecker:48,96,0
+721:wrench:176,96,0
+722:zebra:48,224,0
+723:zucchini:176,224,0

datumaro/datumaro/plugins/voc_format/converter.py

@@ -110,8 +110,8 @@ class _Converter:
         self._images_dir = images_dir
 
     def get_label(self, label_id):
-        return self._strip_label(self._extractor. \
+        return self._extractor. \
-            categories()[AnnotationType.label].items[label_id].name)
+            categories()[AnnotationType.label].items[label_id].name
 
     def save_subsets(self):
         subsets = self._extractor.subsets()
@@ -207,7 +207,7 @@ class _Converter:
 
                         obj_elem = ET.SubElement(root_elem, 'object')
 
-                        obj_label =  self.get_label(obj.label)
+                        obj_label = self.get_label(obj.label)
                         ET.SubElement(obj_elem, 'name').text = obj_label
 
                         if 'pose' in attr:
@@ -357,8 +357,7 @@ class _Converter:
                 for item, item_labels in class_lists.items():
                     if not item_labels:
                         continue
-                    item_labels = [self._strip_label(self.get_label(l))
+                    item_labels = [self.get_label(l) for l in item_labels]
-                        for l in item_labels]
                     presented = label in item_labels
                     f.write('%s % d\n' % (item, 1 if presented else -1))
 
@@ -410,23 +409,37 @@ class _Converter:
         path = osp.join(self._save_dir, VocPath.LABELMAP_FILE)
         write_label_map(path, self._label_map)
 
-    @staticmethod
-    def _strip_label(label):
-        return label.lower().strip()
-
     def _load_categories(self, label_map_source=None):
         if label_map_source == LabelmapType.voc.name:
-            # strictly use VOC default labelmap
+            # use the default VOC colormap
             label_map = make_voc_label_map()
 
-        elif label_map_source == LabelmapType.source.name:
+        elif label_map_source == LabelmapType.source.name and \
-            # generate colormap from the input dataset
+                AnnotationType.mask not in self._extractor.categories():
+            # generate colormap for input labels
             labels = self._extractor.categories() \
                 .get(AnnotationType.label, LabelCategories())
             label_map = OrderedDict()
             label_map['background'] = [None, [], []]
             for item in labels.items:
-                label_map[self._strip_label(item.name)] = [None, [], []]
+                label_map[item.name] = [None, [], []]
+
+        elif label_map_source == LabelmapType.source.name and \
+                AnnotationType.mask in self._extractor.categories():
+            # use source colormap
+            labels = self._extractor.categories()[AnnotationType.label]
+            colors = self._extractor.categories()[AnnotationType.mask]
+            label_map = OrderedDict()
+            has_black = False
+            for idx, item in enumerate(labels.items):
+                color = colors.colormap.get(idx)
+                if idx is not None:
+                    if color == (0, 0, 0):
+                        has_black = True
+                    label_map[item.name] = [color, [], []]
+            if not has_black and 'background' not in label_map:
+                label_map['background'] = [(0, 0, 0), [], []]
+                label_map.move_to_end('background', last=False)
 
         elif label_map_source in [LabelmapType.guess.name, None]:
             # generate colormap for union of VOC and input dataset labels
@@ -436,11 +449,10 @@ class _Converter:
             source_labels = self._extractor.categories() \
                 .get(AnnotationType.label, LabelCategories())
             for label in source_labels.items:
-                label_name = self._strip_label(label.name)
+                if label.name not in label_map:
-                if label_name not in label_map:
                     rebuild_colormap = True
-                if label.attributes or label_name not in label_map:
+                if label.attributes or label.name not in label_map:
-                    label_map[label_name] = [None, [], label.attributes]
+                    label_map[label.name] = [None, [], label.attributes]
 
             if rebuild_colormap:
                 for item in label_map.values():
@@ -469,27 +481,26 @@ class _Converter:
         self._label_id_mapping = self._make_label_id_map()
 
     def _is_label(self, s):
-        return self._label_map.get(self._strip_label(s)) is not None
+        return self._label_map.get(s) is not None
 
     def _is_part(self, s):
-        s = self._strip_label(s)
         for label_desc in self._label_map.values():
             if s in label_desc[1]:
                 return True
         return False
 
     def _is_action(self, label, s):
-        return self._strip_label(s) in self._get_actions(label)
+        return s in self._get_actions(label)
 
     def _get_actions(self, label):
-        label_desc = self._label_map.get(self._strip_label(label))
+        label_desc = self._label_map.get(label)
         if not label_desc:
             return []
         return label_desc[2]
 
     def _make_label_id_map(self):
         source_labels = {
-            id: self._strip_label(label.name) for id, label in
+            id: label.name for id, label in
             enumerate(self._extractor.categories().get(
                 AnnotationType.label, LabelCategories()).items)
         }

datumaro/datumaro/util/mask_tools.py

@@ -9,6 +9,12 @@ from datumaro.util.image import lazy_image, load_image
 
 
 def generate_colormap(length=256):
+    """
+    Generates colors using PASCAL VOC algorithm.
+
+    Returns index -> (R, G, B) mapping.
+    """
+
     def get_bit(number, index):
         return (number >> index) & 1
 

datumaro/tests/test_voc_format.py

@@ -708,7 +708,7 @@ class VocConverterTest(TestCase):
                 SrcExtractor(), VocConverter(label_map='guess'),
                 test_dir, target_dataset=DstExtractor())
 
-    def test_dataset_with_source_labelmap(self):
+    def test_dataset_with_source_labelmap_undefined(self):
         class SrcExtractor(TestExtractorBase):
             def __iter__(self):
                 yield DatasetItem(id=1, annotations=[
@@ -718,7 +718,7 @@ class VocConverterTest(TestCase):
 
             def categories(self):
                 label_cat = LabelCategories()
-                label_cat.add('Label_1') # should become lowercase
+                label_cat.add('Label_1')
                 label_cat.add('label_2')
                 return {
                     AnnotationType.label: label_cat,
@@ -727,7 +727,7 @@ class VocConverterTest(TestCase):
         class DstExtractor(TestExtractorBase):
             def __iter__(self):
                 yield DatasetItem(id=1, annotations=[
-                    Bbox(2, 3, 4, 5, label=self._label('label_1'),
+                    Bbox(2, 3, 4, 5, label=self._label('Label_1'),
                         id=1, group=1, attributes={
                             'truncated': False,
                             'difficult': False,
@@ -746,7 +746,7 @@ class VocConverterTest(TestCase):
             def categories(self):
                 label_map = OrderedDict()
                 label_map['background'] = [None, [], []]
-                label_map['label_1'] = [None, [], []]
+                label_map['Label_1'] = [None, [], []]
                 label_map['label_2'] = [None, [], []]
                 return VOC.make_voc_categories(label_map)
 
@@ -755,6 +755,52 @@ class VocConverterTest(TestCase):
                 SrcExtractor(), VocConverter(label_map='source'),
                 test_dir, target_dataset=DstExtractor())
 
+    def test_dataset_with_source_labelmap_defined(self):
+        class SrcExtractor(TestExtractorBase):
+            def __iter__(self):
+                yield DatasetItem(id=1, annotations=[
+                    Bbox(2, 3, 4, 5, label=0, id=1),
+                    Bbox(1, 2, 3, 4, label=2, id=2),
+                ])
+
+            def categories(self):
+                label_map = OrderedDict()
+                label_map['label_1'] = [(1, 2, 3), [], []]
+                label_map['background'] = [(0, 0, 0), [], []] # can be not 0
+                label_map['label_2'] = [(3, 2, 1), [], []]
+                return VOC.make_voc_categories(label_map)
+
+        class DstExtractor(TestExtractorBase):
+            def __iter__(self):
+                yield DatasetItem(id=1, annotations=[
+                    Bbox(2, 3, 4, 5, label=self._label('label_1'),
+                        id=1, group=1, attributes={
+                            'truncated': False,
+                            'difficult': False,
+                            'occluded': False,
+                        }
+                    ),
+                    Bbox(1, 2, 3, 4, label=self._label('label_2'),
+                        id=2, group=2, attributes={
+                            'truncated': False,
+                            'difficult': False,
+                            'occluded': False,
+                        }
+                    ),
+                ])
+
+            def categories(self):
+                label_map = OrderedDict()
+                label_map['label_1'] = [(1, 2, 3), [], []]
+                label_map['background'] = [(0, 0, 0), [], []]
+                label_map['label_2'] = [(3, 2, 1), [], []]
+                return VOC.make_voc_categories(label_map)
+
+        with TestDir() as test_dir:
+            self._test_save_and_load(
+                SrcExtractor(), VocConverter(label_map='source'),
+                test_dir, target_dataset=DstExtractor())
+
     def test_dataset_with_fixed_labelmap(self):
         class SrcExtractor(TestExtractorBase):
             def __iter__(self):