Attribute Datasets
Attribute is an important type of semantic properties shared among different objects or activities. It is a representation in a higher level than the raw feature representation directly extracted from images or videos. In recent years, several attribute datasets are used by various researchers for the study of utilizing attributes for different vision applications.
1. aPascal & aYahoo Datasets
Application: Object Recognition
Attributes: 64 types of binary attributes annotated for each object sample of the aPascan train and test sets, and the aYahoo test set.
Object Types in aPascal: 20 types of objects for PASCAL VOC2008 Challenge, i.e. people, bird, cat, cow, dog, horse, sheep, aeroplane, bicycle, boat, bus, car, motorbike, train, bottle, chair, dining table, potted plant, sofa and tv/monitor.
Object Types in aYahoo: wolf, zebra, goat, donkey, monkey, statue of people, centaur, bag, building, jet ski, carriage, and mug.
Examples of Object & Attributes:
Acknowledgements: A. Farhadi, I. Endres, D. Hoiem, and D.A. Forsyth, "Describing Objects by their Attributes", CVPR 2009.
2. Animals with Attributes (AWA) Dataset
Application: Object (Animal) Recognition
Attributes: 85 numeric attribute values for each of the 50 animal classes.
Animal Types for Training: antelope, grizzly bear, killer whale, beaver, dalmatian, horse, german shepherd, blue whale, siamese cat, skunk, mole, tiger, moose, spider monkey, elephant, gorilla, ox, fox, sheep, hamster, squirrel, rhinoceros, rabbit, bat, giraffe, wolf, chihuahua, weasel, otter, buffalo, zebra, deer, bobcat, lion, mouse, polar bear, collie, walrus, cow, dolphin.
Animal Types for Testing: chimpanzee, giant panda, leopard, persian cat, pig, hippopotamus, humpback whale, raccoon, rat, seal.
Acknowledgements: C. H. Lampert, H. Nickisch, and S. Harmeling, "Learning To Detect Unseen Object Classes by Between-Class Attribute Transfer", CVPR 2009.
3. Public Figures Face Database (PubFig)
Application: Face Recognition & Face Verification
Attributes: 73 attribute values for each image in PubFig. Examplar facial attributes are "Middle Aged", "Blond Hair", "Big Lips" etc.
Public Figures: 60 individuals in Development set, 140 individuals in Evaluation Set.
Examples of Face & Attributes:
Acknowledgements: Neeraj Kumar, Alexander C. Berg, Peter N. Belhumeur, and Shree K. Nayar, "Attribute and Simile Classifiers for Face Verification", ICCV 2009.
4. Datasets for Human Action Recognition with Attributes
Liu et al. (CVPR2011) proposed a model for recognizing human actions by attributes. In this work, attribute vectors for each action class are defined for different existing human action datasets including UIUC action dataset, Weizmann dataset, KTH dataset and Olympic Sports Dataset. Please see our webpage on action datsets (Link) for more details on these existing human action datasets.
Application: Human Action Recognition
Attributes on UIUC Dataset: 22 action attributes are manually defined for each of the 14 human action classes.
Attributes on Mixed Action Dataset: 34 action attributes are manually defined for each of the 21 human action classes from the mixed UIUC Action, Weizmann, and KTH datasets.
Attributes on Olympic Sports Dataset: 41 action attributes are manually defined for each of the 16 human action classes.
Attribute Definitions: Please see the Technical Report (PDF) by Liu et al.
Acknowledgements: J. Liu, B. Kuipers, S. Savarese, "Recognizing Human Actions by Attributes", CVPR 2011.
5. SUN Attribute Dataset
Application: Scene Attribute Recognition
Attributes: 102 scene attributes are defined for each of the 14,340 scene images.
Scene Types: 717 scene categories
Examples of Images:
Acknowledgements: Genevieve Patterson, James Hays, "SUN Attribute Database: Discovering, Annotating, and Recognizing Scene Attributes", CVPR 2012.
6. Datasets with Relative Attributes
Instance-level relative attributes are defined by A. Kovashka, D. Parikh and K. Grauman in several datasets including Shoes (i.e. Attribute Discovery Dataset), OSR and PubFig datasets.
Application: Scene Recognition, WhittleSearch, etc.
Shoes: 14,658 shoe images augmented with 10 instance-level relative attributes: pointy at the front, open, bright in color, high at the heel, covered with ornaments, shiny, long on the leg, formal, sporty, feminine. (Link)
OSR: 2,688 images with 6 instance-level relative attributes. (Link)
PubFig: 772 images selected from PubFig dataset with 11 instance-level relative attributes. (Link)
[1] D. Parikh and K. Grauman, "Relative Attributes", ICCV 2011.
[2] A. Kovashka, D. Parikh and K. Grauman, "WhittleSearch: Image Search with Relative Attribute Feedback", CVPR 2012.
7. Clothing Attribute Dataset
Appliation: Clothing Attribute Recognition
Attributes: 26 ground truth clothing attributes collected using Amazon Mechanical Turk. Examplar clothing attributes are "Necktie", "Collar" and "Spotted Pattern".
Images: 1856 clothing images.
Acknowledgements: H. Chen, A. Gallagher, B. Girod, "Describing Clothing by Semantic Attributes", ECCV 2012.
8. Social Activity Dataset with Attributes
Application: Social Activity Recognition
Attributes: 69 attributes that can be broken into five broad classes: actions, objects, scenes, sounds, and camera movement.
Social Activity Types: 8 classes of social activities including birthday party, graduation party,music performance, non-music performance, parade, wedding ceremony, wedding dance and wedding reception.
Videos: This dataset provides Youtube IDs of videos coming from CCV dataset. Extracted activity features are also provided.
Acknowledgements: Y. Fu, T. Hospedales, T. Xiang and S. Gong, "Attribute Learning for Understanding Unstructured Social Activity", ECCV 2012.
9. Names 100 Dataset
Application: Gender Recognition & Age Classification
Attributes: the classification scores from each pairwise name classifier of 100 common first names used in the United States.
Images: 80,000 unconstrained face images, including 100 popular names and 800 images per name.
Acknowledgements: uizhong Chen, Andrew Gallagher, and Bernd Girod, "What's in a Name: First Names as Facial Attributes", CVPR 2013.
10. Caltech-UCSD Birds-200 & Caltech-UCSD Birds-200-2011 Datasets
Caltech-UCSD Birds-200 (CUB-200) is an image dataset with photos of 200 types of bird species. Caltech-UCSD Birds-200-2011 (CUB-200-2011) is an extended version of of the CUB-200 dataset. Here, we list the details of the extended CUB-200-2011 dataset.
Attributes: 312 binary attributes per image.
Bird Types: 200 categories of bird species with 11,788 images.
Other Annotations: 15 part locations and 1 bounding box per image.
Bird Image Examples:
[1] Welinder P., Branson S., Mita T., Wah C., Schroff F., Belongie S., Perona, P. "Caltech-UCSD Birds 200". California Institute of Technology. CNS-TR-2010-001. 2010.
[2] Wah C., Branson S., Welinder P., Perona P., Belongie S. "The Caltech-UCSD Birds-200-2011 Dataset". Computation & Neural Systems Technical Report, CNS-TR-2011-001.
11. Outdoor Scene Attribute (SceneAtt) Dataset
The SceneAtt dataset is used for studying shared attribute models, and scene spatial configurations. The dataset is selected from LabelMe Outdoor dataset and SUN Attribute dataset.
Attributes: 17 noun attribute and 30 noun+adjective attribute pairs.
Images: 1226 images of 256*256 pixels in size.
Acknowledgements: S.Wang, J.Joo, Y.Wang and S.C.Zhu. Weakly Supervised Learning for Attribute Localizaiton in Outdoor Scenes. CVPR, 2013.
12. Attributes of People Dataset
Application: person attribute recognition
Attributes: is male, has long hair, has glasses, has hat, has t-shirt, has long sleeves, has shorts, has jeans, has long pants.
Images: 4013 training images and 4022 testing images.
Annotations: person bounding box and 9 ground truth attributes fro each image.
Related Poselet Website: Link
Acknowledgements: L. Bourdev, S. Maji, and J. Malik. Describing People: Poselet-Based Approach to Attribute Classi?cation. In ICCV, 2011.
13. Database of Human Attributes (HAT)
Application: person attribute recognition
Attributes & Images: annotations for 27 attributes of 9344 images
Acknowledgements: G. Sharma and F. Jurie, Learning discriminative spatial representation for image classification, British Machine Vision Conference, 2011.