1. Weizmann Human Behavior Library
This database includes a total of 90 videos, each of which was performed by 9 individuals with 10 different actions (Bend, Jack, Jump, Pjump, run, side, skip, walk, Wave1,wave2). The video's background, perspective, and camera are still. And the database provides a well-labeled foreground contour video. However, the correct rate of this database has reached 100%, and now the article basically no one used it. : http://www.wisdom.weizmann.ac.il/~vision/SpaceTimeActions.html
2. KTH Human Behavior DatabaseThe database consists of 6 classes of behaviors (walking, jogging, running, boxing, hand waving, hand clapping), performed by 25 different people, in four scenarios, with 599 segments in each video. The background is relatively static, except that the lens pulls away, the motion of the camera is relatively slight. This database is now benchmark, the correct rate needs to reach more than 95.5% to be able to send the article. : http://www.nada.kth.se/cvap/actions/
3. Inria xmax Multi-view video libraryThe database was obtained from five perspectives, with a total of 11 people performing 14 acts. A total of 5 cameras are installed in four directions indoors and overhead. In addition, the background and illumination are basically the same. : HTTP://4DREPOSITORY.INRIALPES.FR/PUBLIC/VIEWGROUP/6
4. UCF Sports DatabaseThe video includes 150 videos on sports, a total of 13 actions. The laboratory uses a cross-validation method. 2011 CVPR have used this database, the correct rate to reach 87% to send the article. : http://vision.eecs.ucf.edu/data.html
5. Hollywood Human Behavior LibraryThe database includes 8 types of behavior. These are all fragments of the movie. : http://www.di.ens.fr/~laptev/actions/hollywood2/
6. Olympic Sports DataSetThe database has 16 behaviors, 783 videos. The correct rate now is around 75%. : http://vision.stanford.edu/Datasets/OlympicSports/
7. UIUC Action DataSetThis database has been done 98%, it is recommended not to do it. : http://vision.cs.uiuc.edu/projects/activity/
8. Trecvid Video Library
Human Visual Database