Skip to content

LUNA PLATFORM v.5.67.0#

LP changes

  • The ability to decode video on GPU in video analytics request has been added.

    GPU decoding is enabled by specifying "gpu" in the "decoder_device_class" setting of the "LUNA_REMOTE_SDK_VIDEO_SETTINGS" section.

  • SDK has been updated to version 5.21.0.

    In this version LUNA PLATFORM:

    • HumanDetector has been updated to version v6.
    • DeepFakeDetector has been updated to v3.

    Important: The default threshold for redetection of FaceDetV3 detector has been updated from "0.3" to "0.5". When LUNA PLATFORM is updated, the threshold for redetection in the "redetect_score_threshold" setting of the "LUNA_REMOTE_SDK_FACE_DETECTOR_SETTINGS" section will not be updated automatically. You must manually update the threshold to "0.5" to use the default redetection logic. If a custom value was previously set, you must update the threshold value yourself.

    When installing LUNA PLATFORM from scratch, the "redetect_score_threshold" value will also default to "0.3". It is necessary to use the dump file "platform_settings.json" from the distribution package, or update the threshold value after installation. In the next release, when installing from scratch, the value will be automatically set to "0.5".

  • A mechanism for specifying trusted detections has been added, enabling you to explicitly specify which detections do not need to be redetected.

    It is assumed that trusted detections were obtained using VisionLabs algorithms. Marking a third-party detection as trusted may affect the evaluation results.

    You can indicate whether the detection is trusted using the "application/json" scheme (the "trusted_detections" parameter) or the "multipart/form-data" scheme (the "X-Luna-Trusted-Detections" header) in the following requests, where detections are explicitly specified instead of images:

  • An optional parameter origin_bounding_box has been added to the "generate stream events (beta)" request with specifying the source image (sources > source_type > raw_image), allowing to specify the coordinates of the bbox with a face or body in the coordinate system of the source image stored in the image_origin field of the generated event.

    This enables you to take into account various use cases, such as processing cropped or other modified images. Thus, the old bounding_box parameter is responsible for the coordinates in the image on which the estimates are performed, while the new origin_bounding_box parameter is responsible for the coordinates in the source image.

    This also enables more flexibility in image processing, including the ability to crop or resize the source image before applying the bbox.

    Also, the parameter sources > source > face/body > bounding_box, available when specifying detections in a request (sources > source_type > detections), has been renamed to origin_bounding_box.

  • The support for setting the DROI relative to the source frame has been added.

    If ROI is intended to optimize the estimation of the corresponding features, then DROI works as a filter after the estimation has been completed and is intended to implement business logic. You can use DROI both together with ROI and separately. For example, if after processing by the ROI area the number of people in the frame was equal to N, then after additional filtering by the DROI area the number of people in the frame can be reduced to M.

    The DROI on the source frame is specified by the following parameters:

    • "area" — Geometry of the region of interest. The parameter is represented as an array of polygons. Each polygon is represented by an array of objects, where each object contains the x and y coordinates of the polygon's vertex. For example, you can define a triangular area of interest.
    • "mode" – mode of specifying “x”, “y”. Two modes are available:
      • "abs" — Parameters "x", "y", "width" and "height" are set in pixels.
      • "percent" — Parameters "x", "y", "width" and "height" are set as percentages of the current frame size.
    • "form" — Format of the region of interest. In the current implementation there is only one possible value, "common". Future implementations will support other formats in addition to polygons.

  • The ability to set the time to live of task results has been added to all tasks.

    TTL is set in result_storage_policy > ttl parameter in task or schedule creation requests.

    It is not possible to add TTL of task results to already created and executed tasks.

    You can add the TTL of task results to an already created schedule by using the "replace tasks schedule" request. TTL of task results will not be applied to tasks created or running at the time of the request.

  • Added automatic sending of the header Set-Cookie: LUNA_AUTH_TOKEN=""; Path=/; Max-Age=0 to the client application when it receives a 401 error from the API service due to a rotten authorization token placed in the Cookie.

LP fixed errors

  • Fixed migration of Configurator settings with revision c51cdfac.

    This migration failed if the bucket field was not present in the S3 setting of the Image Store service.

    The bucket field was removed from the Image Store "S3" section in version 5.45.4.

  • Removed unnecessary authorization for "get version", "get health", "set login cookie" and "clear login cookie" requests.