Center for Biometrics and Security Research
Home | News | Team | Projects | Research | Standards | Demos | Databases | Parteners | Contact      ÖÐÎÄ
SEARCH£º        
 
CASIA action database for recognition

    CASIA action database is a collection of sequences of human activities captured by video cameras outdoors from different angle of view. There are 1446 sequences in all containing eight types of actions of single person (walk, run, bend, jump, crouch, faint, wander and punching a car) performed each by 24 subjects and seven types of two person interactions (rob, fight, follow, follow and gather, meet and part, meet and gather, overtake) performed by every 2 subjects.


Description:
    Single person action:
        Walk: One subject walking along the road.
        Run: One subject running along the road.
        Bend: One subject bending his/her body while walking.
        Jump: One subject jumping along the road.
        Crouch: One subject crouching once a time while walking along the road.
        Faint: One subject falling down to the ground while walking along the road.
        Wander: One subject wandering around.
        Punching a car: One subject punching a car.
    Interaction:
        Rob: One subject robing another one.
        Fight: Two subjects fighting with each other.
        Follow: One subject Following another one till the end.
        Follow and gather: One subject Following another one and then walking together with the other to the end.
        Meet and part: Two subjects meeting each other and then departing.
        Meet and gather: Two subjects meeting each other and then walking together till the end.
        Overtake: One subject overtaking another one.


Sample Image (original size 320*240):

sample

¡¡¡¡

    All video sequences were taken simultaneously with static three non-calibrated cameras from different angle of view (horizontal view, angle view and top down view) with the frame rate of 25fps and were compressed with the huffyuv code in avi format. The original spatial resolution of the sequences was resized to 320*240. The duration of the sequences lasts from 5 seconds to 30 seconds due to different kind of actions.

Sample sequences (resized):

Angle view:

seqSample

Horizontal view:

seq2Sample

Top down view:

seq3Sample

 

    The avi files are named with items, like [view angle]_[person label]_[action]_[action count].avi£¬For example, angleview_p01_walk_a1.avi , representing angle of view, persons¡¯ ID, action category and action No. respectively.


The database archive structure:


Download:

File Name List (TXT)

To apply for the database£¬please follow the steps below£º

1. download and print the document Agreement for using CASIA database

2. sign the agreement£»

3. send the agreement copy to us by E-mail: casia_ar@cripac.ia.ac.cn.

4. If your application is passed, you can download the dataset from our ftp server.


Publications

  1. Ying Wang, Kaiqi Huang, Tieniu Tan.Human Activity Recognition Based on R Transform. Computer Vision and Pattern Recognition, 2007. CVPR '07
  2. Zhang Zhang, Kaiqi Huang, Tieniu Tan, Multi-Thread Parsing for Recognizing Complex Events in Videos. European Conference on Computer Vision, 2008. ECCV'08.
  3. Zhang Zhang, Kaiqi Huang, Tieniu Tan, Liangsheng Wang, Trajectory Series Analysis based Event Rule Induction for Visual Surveillance. Computer Vision and Pattern Recognition, 2007. CVPR'07.
¡¡  Introduction
¡¡  Iris Databases
¡¡  Gait Databases
¡¡  HFB Face Databases
¡¡  NIR Face Databases
¡¡  BIT Face Databases
¡¡  Fingerprint Databases
¡¡  Handwriting Databases
¡¡  Action Databases
¡¡  Palmprint Databases
¡¡  Multi-spectral Palmprint Databases
Copy rigth All right reserved 2005 Center for Biometrics and Security Research
Center for Biometrics and Security Research 12th Floor,Institute of Automation chinese Academy of Sciences
P.O.Box2728Beijing 100080 P.R.China
Tel:010-62632259 Fax:010-62632259 E-MAIL:hjzhang@cbsr.ia.ac.cn