Skip to content

LUNA PLATFORM v.5.94.0#

LP changes

  • SDK has been updated to version 5.26.0.

    In this version of LUNA PLATFORM:

    • DeepFake estimator updated to version 7;
    • OneShotLiveness estimator updated to version 9;
    • the face descriptor neural network of version 54 is no longer supported;
    • the 65th neural network model for extracting facial biometric templates has been added to the Remote SDK and Video Agent containers.

    Note: Note that the added network is already in the containers of the specified services and does not need to be added manually.

    Important: If the 54th neural network model is used at the time of the update (the "DEFAULT_FACE_DESCRIPTOR_VERSION" setting), then without the following actions the Remote SDK service will not start:

    • Perform the "Additional extraction" task before the update and then specify the required version in the "DEFAULT_FACE_DESCRIPTOR_VERSION" before the Remote SDK launch. Thus, you can use the previous neural network version.
    • Specify the new version in "DEFAULT_FACE_DESCRIPTOR_VERSION" before the Remote SDK launch. Already created descriptors can no longer be used if the "Additional extraction" task was not performed.

  • Now Python Matcher service has been classified as optional and is regulated within the group of parameters "ADDITIONAL_SERVICE_USAGE".

    To enable matching requests, at least one of the services, Python Matcher or Python Matcher Proxy, must remain enabled; if both services are disabled at the same time, this will result in the following consequences:

    • It will be impossible to use the matching policy "match_policy" of handler.
    • It will not be possible to execute requests to the resource "/matcher".
    • Creation of Clustering, ROC-curve calculating and Cross-matching tasks will not be available.

    In the request "get platform features" a new component "human_matching" has also been added, the status of which depends on the state of the Python Matcher and Python Matcher Proxy services.

    The component takes the value "true" if at least one of the services is enabled, and "false" if both services are disabled.

  • Support for working with descriptors has been added to general events.

    In the request create new general event a new field descriptors has been added, which represents a set of descriptors of different types.

    A new request "get general event descriptors" has also been added that allows you to get general event descriptors.

    Note that not only face or body descriptors can be written to general events. The main thing is that descriptors should be in SDK format.

  • To improve the speed of query execution, the tables general_event_location and general_event of the Events service have been merged into one.

    Now the location information, namely city, area, district, street, house_number, geo_position, is stored in the table general_event.

    Despite these changes, the structure of server requests and responses has not changed.

  • The process of writing data to the "general_events" table has been modified to optimize the database.

    Previously, the data of the fields "event_id", "track_id", "stream_id", "source" and "location" was written twice: in the "event" field in JSON format and as independent fields in the table.
    Now the duplication is eliminated and the data is written exclusively to separate fields of the "general_events" table.

  • The ability to use the value last_processed_frame_time of the stream as a filter for the request get streams logs has been added.

    last_processed_frame_time - the last processed frame time since stream start. If no frames were processed, the parameter takes the value null.

  • The ability to specify a custom schema for PostgreSQL databases generated using the Storages utility has been added.

    The scheme is specified by passing the LUNA_PG_SCHEMA environment variable when starting the Storages container to prepare the environment (the prepare argument).

    LUNA_PG_SCHEMA specifies the schema name of PostgreSQL when creating a table. The default value is luna. If LUNA_PG_SCHEMA is not present, this schema will be used by default for new tables, making it the default schema for the corresponding user.

    Previously, the user luna was only able to work with the schema named public.

    This allows you to better organize your data and isolate it for different users.

    Important: If the schema name is different from the user name, the table will be created in the public schema.

    See the command example in the section "Setting up a custom schema for databases".

  • Functionality has been added that allows the use of Lambda for distributing static content, such as user interfaces, HTML pages, JavaScript scripts, or images.

    To do this, you need to place the files in the "static" directory in the archive with lambda.

    See an example of lambda working with files in the section "Standalone lambda examples".

  • You can now add custom endpoints to any type of lambda.

    For example, you could create a /parameters/{UUID} endpoint to allow you to store and validate unique parameters, or add other endpoints to handle specific requests within the lambda.

    See details and examples in the section "Additional routes".

LP fixed errors

  • The human tracking analytics specification has been updated to include the "image_retain_policy" parameter, which was previously not exposed.

  • A missing query has been added to the API service specification stream events ws handshake.

  • Fixed an error that occurred when sending a "get statistics on events" request using the count aggregator for meta content targets with a non-numeric data type.

    Previously, in this case, getting event statistics resulted in an error with code 10020 (incompatibility of user-defined filters).

  • Fixed a memory leak in the Video Agent service that occurred during video decoding on the GPU.