logo logo

Configure camera

You can easily edit the camera and its configuration. The camera configuration is accessible in the right panel after clicking on the camera name.

The camera configuration right sliding panel for the camera with name "Entry to hall".

Camera settings

The SmartFace Station allows you to configure camera settings. These settings directly affecting the processing of video streams done by SmartFace Platform.

To edit the camera settings:

  1. Click on Configuration in the left menu panel.

  2. Go to Camera settings section.

  3. Click on the camera name.

You may edit these parameters:

Parameter nameParameter descriptionDefault valueValue intervalNote
RTSP / VIDEO SOURCEA RTSP URL to the camera stream or video file name with full path for file location
OS SERVICE NAMEThe camera service name<assigned by the system>read only information
MAXIMUM TRACKING FACESThe maximum number of faces detected on one frame20Setup this value according to your use-case. Lowering this number, you can significantly decrease overall detection speed.
MIN FACE SIZEThe minimum face size necessary for detection in pixels35
MAX FACE SIZEThe maximum face size necessary for detection in pixels600
CONFIDENCE THRESHOLDThe confidence threshold for face detection450
FACE SAVE STRATEGY

The strategy for saving data from all detections, you can choose from:

  • All- All faces for the detected person are stored.
  • FirstFace - Only the first detected face of a person is stored.
  • BestFace - Only the best face for the detected person is stored. Which face is the best is decided when the person is lost from tracking.
  • FirstFace, BestFace - The first and the best face for the detected person are stored.
  • MatchedOnly - Only faces which are matched against a watchlist member are stored.

FirstFace, BestFaceSet this wisely. Strategy to save all detected objects can generate huge amount of data.
FACE DETECTOR RESOURCEChoose between these resources for detection:

  • None - no face detection enabled
  • CPU - face detection using NN with balanced mix of speed vs. accuracy, performed directly in the camera process on CPU
  • CPU Remote - face detection using NN with balanced mix of speed vs. accuracy, performed remotely in Detector process on CPU
  • Accurate CPU - face detection using NN that is more accurate but slower, performed directly in the camera process on CPU
  • Accurate CPU remote - face detection using NN that is more accurate but slower, performed remotely in Detector process on CPU
  • GPU - face detection using NN with balanced mix of speed vs. accuracy, performed directly in the camera process on GPU
  • GPU Remote - face detection using NN with balanced mix of speed vs. accuracy, performed remotely in detector process on GPU
  • Accurate GPU - face detection using NN that is more accurate but slower, performed directly in the camera process on GPU
  • Accurate GPU remote - face detection using NN that is more accurate but slower, performed remotely in detector process on GPU
  • Any Remote - face detection using NN with balanced mix of speed vs. accuracy, performed remotely in detector process on any resource (GPU or CPU) according to their current utilization.
  • Accurate Any remote - face detection using NN that is more accurate but slower, performed remotely in detector process on any resource (GPU or CPU) according to their current utilization.

CPU
TEMPLATE GENERATOR RESOURCEChoose between these resources for template extraction:

  • CPU - to perform the template extraction on CPU resource.
  • GPU - to perform the template extraction on GPU resource. GPU must be present in the system.
  • ANY - to perform the template extraction on any resource (GPU or CPU) according to their current utilization. GPU must be present in the system.

CPU
FACE DISCOVERY FREQUENCYThe frequency of face detection in milliseconds500
FACE EXTRACTION FREQUENCYThe frequency of face extraction in milliseconds500
PEDESTRIAN DETECTOR RESOURCE

Turn on pedestrian detection on selected camera. Choose which detector will be used for pedestrian detection.

You can choose from:

  • None - no pedestrian detection enabled
  • CPU Remote - use special Innovatrics NN to perform the detection on CPU
  • GPU Remote - use special Innovatrics NN to perform the detection on GPU. GPU must be present in the system.
  • Any Remote - use special Innovatrics NN to perform the detection on any resource (GPU or CPU) according to their current utilization. GPU must be present in the system.
  • BodyParts - use body-parts (CPU) detector to determine where is the pedestrian is located on image

None
MIN PEDESTRIAN SIZEThe minimum pedestrian size necessary for detection in pixels8020 ↔︎ 10 000
MAX PEDESTRIAN SIZEThe maximum pedestrian size necessary for detection in pixels2 00021 ↔︎ 10 000
MAX PEDESTRIANSThe maximum number of pedestrians detected on one frame201 ↔︎ 1 000
PEDESTRIAN CONFIDENCE THRESHOLDThe confidence threshold for pedestrian detection2 5000 ↔︎ 10 000
SPOOF DETECTOR RESOURCE

Choose liveness approach for spoof detector.

You can choose from:

  • None - no spoof detection enabled
  • Distant CPU - NN based liveness for access control use-case executed on CPU
  • Distant GPU - NN based liveness for access control use-case executed on GPU
  • Distant ANY - NN based liveness for access control use-case executed on any resource (GPU or CPU) according to their current utilization.
  • Nearby CPU - NN based liveness for selfie use-case executed on CPU
  • Nearby GPU- NN based liveness for selfie use-case executed on GPU
  • Nearby ANY - NN based liveness for selfie use-case executed on any resource (GPU or CPU) according to their current utilization.
  • Body parts - spoof detection based on body pose (hands holding printout of static photo or a tablet)

you can set only one spoof detector resource per camera via UI
DISTANT LIVENESS SCORE THRESHOLDAnything below this value from distant liveness check will be considered as spoof900 ↔︎ 100available only for Distant CPU and Distant GPU SPOOF DETECTOR RESOURCE
NEARBY LIVENESS SCORE THRESHOLDAnything below this value from nearby liveness check will be considered as spoof900 ↔︎ 100available only for Nearby CPU and Nearby GPU SPOOF DETECTOR RESOURCE
FEEDBACK DISPLAYOpens/Copy the [feedback display] (/docs/how-it-works/smartface-station/sf-station-feedback-display/) link for selected camera
PREVIEW ENABLEDEnable preview generation with embedded graphics. The preview is generated as an MPEG1 stream. You can define the stream quality in the PREVIEW QUALITY setting.Disabled
PREVIEW QUALITY

Define the quality of generated stream.

You can choose from:

  • Low (height: 426 px, bit rate: 153,000)
  • Medium (height: 640 px, bit rate: 450,000)
  • High (height: 1280 px, bit rate: 1,400,000)

Low
PREVIEW PROFILES

Define predefined profile what attributes will be displayed near to the detected or identified object.

Four predefined profiles are available:

  • Minimal (bounding box)
  • AC/Security (bounding box, watchlist member name)
  • Retail (bounding box, watchlist member name, age, gender)
  • Full (bounding box, watchlist member name, age, gender, order, facesize / pedestrian size, confidence / pedestrian quality, template quality, yaw, pitch, roll)

You can create your custom profile with different set of attributes via REST API.

Minimal