Page History

Configure the basic settings of the Face Detection module as  module as follows:

1. Go to
the 
2. the Basic settings
 tab
3.  tab on the settings panel of
the 
4. the Face Detection
 object
5.  object.
   Image Modified

6. There are several variations of the face detection operation depending on the setting of the external capture and Face recognition server (1):

   1. External capture is NOT selected, and the detection is NOT added to any channel of the Face recognition server
channels
7. 1. . In this case, faces are captured by Axxon PSIM face capturing, and vectors aren't generated. This option isn't used in Face PSIM.
   2. External capture is NOT selected, but the detection is added to
any
8. 1. some channel of the Face recognition server
channels
9. 1. . In this case, the photos of captured faces are captured and saved
by
10. 1. by Axxon PSIM face capturing, the vectors are generated and saved to the database.
This option does not allow face recognition (the
11. 1. Face recognition, search and visitor counting are available
)
12. 1. with this option.
    2. External capture is selected, but the detection is NOT added to any channel of the Face recognition server
channels
13. 1. . In this case, the photos of faces captured by the face recognition modules (see Structure of the Face PSIM software package) are captured and saved, and the vectors are generated, but the vectors
are not
14. 1. aren't saved to the database. This option is only suitable for use in scripts.

    2. External capture is selected, and the detection is added to any channel of the Face recognition server

channels

15. 1. . In this case, the photos of faces captured by the face recognition modules (

see

16. 1. see Structure of the Face PSIM software package) are captured and saved, and the vectors are generated and saved to the database.

The

17. We recommend using the external capture

is recommended for use

18. . From

the

19. the External capture drop-down list, select the Face recognition server

with the help of which the face capturing

20. that will capture faces with the subsequent generation of a vector (if

the

21. the Create FIR in detection unit checkbox is set on the settings panel of the corresponding Face recognition server

object—see

22. object—see Configuring the Face recognition server object) and saving of photos

will be performed

23. .

    Info
    title Note

Proper configuring of

24. When you use Cognitec face recognizer, you can fine-tune the Face recognition server

selected

25. that you select as an external capture

is performed with the help of the

26. using the Cfgedit.exe

utility

27. utilit (for the detailed information about this utility, see Cfgedit.exe utility for configuring the Cognitec face recognizer).

28. The Maximum performance mode is enabled by default (2). This mode allows improving the quality of face capture and recognition by using all available processor cores and increasing the RAM consumption.
When
29. If you disable this mode
is disabled
30. , the memory consumption
will decrease
31. decreases, but faces can be missed
at
32. when capturing.
    

    Info
    title Note
    When you use VisionLabs face recognizer, the Maximum performance checkbox is cleared automatically, and the setting becomes inactive when you select the Face recognition server in the previous step. This is due to the feature of this recognizer.

33. Set the border of the face search area

border

34. on the video image by specifying the required values in percent in the corresponding fields (3), or by clicking the button (4) and using the interactive interface. After you click the button, the window with the corresponding video camera image

will open

35. opens. To set the border of the face search area

border, click

36. :

    1. Click the Stop video button

at the required time, select

37. 1. .

    2. Select the required area using the left mouse button

, and then click

38. 1. .

    2. Click the OK button.

 The specified face

39. 1. The values of the specified border of the face search area

border values will

40. 1. will be automatically entered in

the

41. 1. the Left, Right, Top and Bottom fields of the Search area border % group.
       Image Modified

       Note
       title Attention!
       Setting the border of the face search area

border

42. 1. is available only

if

43. 1. for the VA

or 

44. 1. and VideoIntellect 1.1 face recognition modules

are used

45. 1. .

46. Set the Scaling and selection slider (5) to the position corresponding to the required scale of the captured face in the frame for recording and highlighting the face with a dynamic frame. The current scale value is displayed in the field above the slider and ranges from 1 (only the captured face is recorded) to 5 (the whole frame is recorded).
47. In the Processing speed, fps field (6), specify how many FPS from the video stream are processed by the Face detection program module. This parameter is necessary for reducing the load on the
Server
48. server.

49. If it is necessary to convert the color video image to a black-and-white image to generate the vectors by the black-and-white frames, then clear the Process color video checkbox (7).

    Note
    title Attention!
    Generation of vectors by black-and-white frames is supported by the VA and VideoIntellect 1.1 face recognition modules only. It reduces the CPU load, insignificantly affecting the face capture and recognition quality.

50. By default, the Face

detection tool

51. detector processes a video with the resolution set for the Camera

 parent

52. parent object. To process a video with 320x240

pixels

53. pixel resolution, set the Process low resolution video checkbox (8).

    Info
    title Note
    Enabling this option reduces the load

of

54. on the

Server’s

55. server processor, but the probability of face capture also reduces.

note

title	Attention!

57. The Process low resolution video setting

is not relevant in case,

58. isn't relevant when a high resolution is set for the Camera parent object. Also, this

settings

59. setting isn't used if the VA or VideoIntellect 1.1 module is

applied for face capture

60. used to capture a face.

61. Set the minimum and maximum width and height of the face on the video image in pixels in the corresponding fields of the Face size, pixels group (9).

    Info
    title Note
    If the specified minimum dimensions differ significantly from the actual ones, then the faces

will not be captured.

62. won't be recognized by the detector. For example, if the video contains large faces (15-30%), then if you set the minimum size to 5%

,

63. and the maximum size to 80%,

 then

64. then the faces

will not

65. won't be detected. However, if you specify a minimum size of 12% and a maximum size of 80%,

the

66. faces will be

captured

67. detected.

68. Thus, if the

detection tool

69. detector is configured to capture small faces, then it cannot detect large ones, and vice versa.

InfotitleNote

70. If all parameters of face

sizes

71. size are set to 0, the Face Detection

tool

72. module uses similar parameters

set on the 

73. specified on the settings panel of the Face recognition server

 object settings panel

74. object (see Configuring the

captured

75. size of a face

size

76. for recognition).

Info

title	Note

When using

77. If you use the VisionLabs face recognition module, the minimum and maximum values for all face

detection tools should

78. detectors must be the same, as the values are set for the entire software module.

79. From the Select GPU drop-down list (10), select a video card for recognition for the VA and VisionLabs modules. The Use GPU checkbox must be set on the settings panel of the Face recognition server object (see Configuring the recognition parameters of captured faces).
    

    Note
    title Attention!
    All GPUs with Pascal architecture are not supported by the VA module; if you select such a graphics card, faces will not be recognized.

80. To save the changes, click the Apply button (
10)
81. 11).

Configuration of the basic settings of the Face Detection module is complete.