NTU Dataset Overview

Dataset Description

The NTU dataset is a collection of videos covering a wide range of human actions. Each action was recorded using three cameras.

List of Recorded Actions

1. Drinking water
2. Eating/snacking
3. Brushing teeth
4. Combing hair
5. Dropping objects
6. Picking up objects
7. Throwing
8. Sitting down
9. Standing up from a seated position
10. Clapping
11. Reading
12. Writing
13. Tearing paper
14. Putting on a jacket
15. Removing a jacket
16. Putting on shoes
17. Removing shoes
18. Wearing glasses
19. Taking off glasses
20. Wearing a hat/cap
21. Removing a hat/cap
22. Cheering
23. Waving
24. Kicking something
25. Putting something into a pocket / taking something out of a pocket
26. Jumping on one foot
27. Jumping
28. Making a phone call / answering a phone call
29. Using a phone/tablet
30. Typing on a keyboard
31. Pointing with a finger
32. Taking a selfie
33. Checking the time (from a watch)
34. Rubbing hands
35. Nodding / bowing
36. Shaking head
37. Rubbing face
38. Saluting
39. Pressing palms together
40. Crossing hands in front (signifying stop)
41. Sneezing / coughing
42. Staggering
43. Falling down
44. Touching head (headache)
45. Touching chest (stomachache/heart pain)
46. Touching back (back pain)
47. Touching neck (neck pain)
48. Nausea or vomiting
49. Using a fan (hand or paper) / feeling hot
50. Hitting / slapping someone
51. Kicking someone
52. Pushing someone
53. Patting someone's back
54. Pointing at someone with a finger
55. Hugging someone
56. Giving something to someone
57. Touching someone's pocket
58. Shaking hands
59. Walking towards each other
60. Walking away from each other

File Naming Convention

Each file is named according to the following information:

• Setup number
• Camera ID
• Performer ID
• Replication number
• Action class label

File Content

Each file contains per‑frame skeletal coordinates. The first three columns of coordinates are retained in the code for training neural networks.