Cross Match task

Cross-matching is comparing two sets of objects based on their descriptors.

To make cross-matching task it needs attribute_ids. If you need to use face attributes for the cross-matching task, set the descriptor type to “face” and use faces from the Luna Faces or events from the Luna Events. If you need to use bodies for the cross-matching task, set the descriptor type to “body” and get the attributes using events from Luna Events. In both cases, only objects with descriptors will be processed. One can optionally specify the match threshold. Also, it needs account_id for task creation.

During the task execution task-worker service creates temporary files with task results in the file system. This is a tradeoff to reduce memory footprint of the service. Although this files are cleaned up during normal operation, make sure that you have enough space for this files. Files are created in OS-specific location (usually its /tmp). This could be overriden with TMPDIR env variable.

Cross Match process

Cross Match is done in several steps:

• collect objects having attribute ids using provided filters

• match every reference object with every candidate object

• match results are sort (lexicographically) and cropped (limit and threshold are applied)

Luna-faces

There are three filters to get faces from service. It can be used separately or together or without any filters.

Filters:

list_id: id of list, which face(s) are linked to
create_time__gte: lower included bound for object create_time
create_time__lt: upper bound for object create_time

Luna-Events

There are some filters to get events from service. It can be used separately or together or without any filters.

Filters:

create_time__lt: upper bound for object create_time
create_time__gte: lower included bound for object create_time
sources: source filter
event_ids: event ids
handler_ids: handler ids
top_matching_candidates_label: top matching candidate label
top_similar_object_ids: top similar object ids
top_similar_external_ids: top similar external ids
top_similar_object_similarity__gte: top similar object similarity lower included bound
top_similar_object_similarity__lt: top similar object similarity upper excluded bound
age__lt: age upper excluded bound
age__gte: age lower included bound
gender: gender filter
emotions: emotions filter
ethnic_groups: ethnic group filter
face_ids: face ids
user_data: user data filter
tags: tags filter
cities: city location filter
areas: area location filter
districts: district location filter
streets: street location filter
house_numbers: house number location filter
geo_position: geo position filter
masks: medical mask state filter
track_ids: track id filter
liveness: liveness state filter