Kernel-Based Multiview Joint Sparse Coding for Image Annotation

Table 1

Comparison of predicted labels with ground-truth annotations for images from Corel5K, IAPRTC12, and MIRFlickr25K datasets. Italic labels are the difference between predicted and ground-truth labels while some of them are also relevant to the image.

Images from Corel5K
Ground truths	Cow, bulls, antlers, elk	Grass, ruins, stone, column	Coral, anemone, ocean, reefs	Grass, bear, polar, tundra	Cars, tracks, turn, prototype
Predicted labels	Meadow, cow, bulls, elk, antlers	Ruins, stone, palace, column, grass	Coral, anemone, ocean, reefs, flowers	Grass, bear, polar, tundra, tree	Cars, tracks, turn, prototype, birds
Images from IAPRTC12
Ground truths	cloth, front, girl, hat, wall	court, man, player, tennis	beach, boat, fog, man, sea, tower	bush, cloud, group, hill, island, lake, people, sky	church, lamp, night, sky, square, street, tower
Predicted labels	front, girl, hat, jacket, wall	court, man, player, tennis, net	beach, boat, fog, man, people	group, hill, island, lake, photo	cathedral, lamp, night, square, street
Images from MIRFlickr25K
Ground truths	animals, bird, , clouds, plant_life, sky, structures, sunset, tree	plant_life, structures, tree,	flower, , night, plant_life	animals, plant_life, structures, tree, water	car, , plant_life, sky, structures, transport, tree
Predicted labels	clouds, plant_life, sky, structures, tree	plant_life, sky, structures, tree,	flower, , plant_life, indoor, night	animals, plant_life, tree, sky, structures	indoor, structures, plant_life, sky, tree