HOHA
From Chorus
Domain | Human actions |
Media | Video, Annotations |
Size | 2.4 GB |
Instances | not known |
File Format | DivX 5 |
Creation Date | 2008 |
Task | action recognition |
Copyright | not known |
URL | http://www.irisa.fr/vista/actions/ |
Domain
- Human Actions and Scenes
- The archive contains the following two directories:
- videoclips
- annotations
- videoclips
Comments
- HOLLYWOOD HUMAN ACTIONS (HOHA)
This archive provides the video samples and annotations used in the experimental section of the paper "Learning realistic human actions from movies" by I. Laptev, M. Marszalek, C. Schmid and B. Rozenfeld,
published in CVPR 2008.
Media (image, video, mixed, …)
- video samples and annotations
Size (no images, in GB, …)
- 2.4Gb
Source (FlickR, Corel)
- Videoclips sources : short sequences from 32 movies:American Beauty, As Good As It Gets, Being John Malkovich, Big Fish, The Big Lebowski, Bringing Out The Dead, The Butterfly Effect, Casablanca, The Crying Game, Dead Poets Society, Double Indemnity, Erin Brockovich, Fargo, Forrest Gump, Gandhi, The Godfather, The Graduate, I Am Sam, Independence Day, Indiana Jones And The Last Crusade, Its A Wonderful Life, Kids, LA Confidential,LOR - Fellowship Of The Ring, Lost Highway, The Lost Weekend, Mission To Mars, The Naked City, The Pianist, Pulp Fiction, Raising Arizona, Reservoir Dogs.
- The content of the annotations directory defines video samples as fragments of the video clips. The fragments are specified by frame ranges. For the automatic training, samples correspond to full clips. For manual annotations, clips could be trimmed or split. Each sample is annotated according to 8 classes: AnswerPhone, GetOutCar, HandShake, HugPerson, Kiss, SitDown, SitUp, StandUp.
Annotation type (free text, structured, …)
Ground truth
Event or project
Task (retrieval, recognition, …)
Format
- The video frames typically consist of 240 lines, the aspect ratios vary.
The videos run at about 24 fps. The clips are encoded using the DivX 5 codec
Quality (resolution)
Creation date
- 2008
Copyright
To cite this database please use:
Ivan Laptev, Marcin Marszałek, Cordelia Schmid and Benjamin Rozenfeld, Learning Realistic Human Actions from Movies, CVPR 2008.