Skip to content

FaceStream v.5.2.0#

Changes

  • The section "INFLUX_MONITORING" has been renamed to "LUNA_MONITORING" in LUNA Streams settings.

    The section has also been updated with the "storage_type" parameter, which defines the type of monitoring data storage. Currently, only the "influx" type is available.

    Important: Since a similar update has been added to LUNA PLATFORM v.5.76.0, the minimum required version of LUNA PLATFORM for working with FaceStream has been updated to v.5.76.0.

  • Detection will no longer be considered the best shot if it does not reach the value specified in the "min_body_size_threshold" parameter.

  • The mechanism for automatic stream restart has been improved for more effective processing of large data volumes.

Fixed errors

  • The missing parameter "preferred_program_stream_frame_width" has been returned in the API V2 of LUNA Streams service.

  • Errors occurring during background tasks of automatic stream restart with a large number of streams have been fixed.

  • In the "create stream V2" request, it is now required to specify at least one item in the "data" > "analytics" section.

  • Various errors occurring when sending reports during a high volume of simultaneous requests with many streams have been fixed.

FaceStream v.5.2.0#

Changes

  • Support for API V2 of the LUNA Streams service has been added.

    The changes in the second version are aimed at using the collaborative mode of operation of the face and body detector. The face and body detector is enabled by simultaneously enabling the "use-face-detector" and "use-body-detector" parameters in the TrackEngine settings.

    API V1 support remains available.

    Main changes in the new version of the API:

    • Setting the address and API version of the LUNA API service has been removed. Now the connection settings with LUNA PLATFORM are specified in the FaceStream settings in the "lunaplatform" section.
    • The "min_score" parameter is divided into two parameters - "min_score_face" and "min_score_body".
    • Parameter "number_of_bestshots_to_send" is divided into two parameters - "face_bestshots_to_send" and "body_bestshots_to_send".
    • Removed all deprecated parameters (for example, "detection_handler").

    The administrator manual has been completely redesigned to use API V2.

    The plugin for proxying LUNA Streams requests via the LUNA API service also supports the new version of the API. See "Proxying requests to LUNA Streams using LUNA API" for details.

    The LUNA Streams user interface does not currently support API V2. Support will be added later.

    Important: At the moment there are restrictions on the operation of the API V2 functionality described below.

    General functionality:

    • Missing parameter "preferred_program_stream_frame_width".
    • It is impossible to specify more than one analytics in an analytics array.

    Collaborative mode of operation of faces and bodies:

    • Section "data" > "analytics" > "primary_track_policy" is not supported.
    • Parameter "data" > "analytics" > "sending" > "full_frame_settings", which determines which source frames will be sent (faces, bodies or faces and bodies), is not supported. Currently, only source frames of faces and bodies are sent.
    • Parameter "data" > "analytics" > "sending" > "bestshot_settings" > "type", which determines which best shots will be sent (faces, bodies or faces and bodies), is not supported. At this time, only the best faces and bodies will be submitted.

    Periodic sending of best photos in shared mode works similarly to working with bodies. This means that the "silent_period" parameter will not work in shared mode.

    A description of the limitations of the collaborative mode is also described in the section "Face and body detection collaborative mode" in the administrator manual.

  • The ability to send Crop images to an external service instead of samples ("warp" format) and modified images ("gost" format) has been added.

    Sending is controlled by the "sending" > "portrait_type" parameter, which now accepts the new value "crop".

    A Crop image is a fragment of an image obtained by expanding and cropping the area around the detected object. The expansion of the area is determined by the new "sending" > "crop_factor" parameter, which increases the dimensions of the original rectangle containing the object. The maximum size of the cropped image is controlled by the new "sending" > "max_crop_size" parameter, which limits the largest side of the area. If the size of the fragment exceeds the specified parameter, the fragment will be scaled proportionally along both coordinates to the desired size. A Crop image allows capturing more context and detail around the object compared to a sample.

    This allows for more flexible image processing. Sending Crop images has several key advantages:

    1. Face coordinates: Along with the Crop image, the coordinates of the face on both the Crop image and the source image are transmitted. This avoids the need to use redetetor, ensuring the presence of the face in the image and saving computational resources.
    2. Additional information: A Crop image contains more context and details around the face, allowing for the estimation of various parameters that cannot be estimated on a biometric. For example, Deepfake estimation, Liveness estimation, etc., can be performed with greater accuracy.
    3. External services: External services and face recognition algorithms often work better with Crop images than with samples. This is because Crop images contain more information and context, improving the accuracy of recognition and analysis.

    Important: Starting from the current version of FaceStream, it is necessary to use LUNA PLATFORM v.5.67.0 and higher. Older versions of LUNA PLATFORM are not supported by the current version of FaceStream.