The Face recognition parameters. The Administrator client
Parameters | A working name of a parameter | A parameter type | A default value | An acceptable range * | A parameter description |
|---|---|---|---|---|---|
Frames per second. | fps | float | actual | [1, 30] | A certain number of frames per second a video is processed. |
A minimum object size. | min_obj_size | float (in pixels) | 30.0 | [0, 512] | A minimum size of the object being recognized (in pixels). The size of the object is determined by the length of its diagonal. |
A minimum object width. | min_obj_rel_width | float | 0 | [0, 1] | The minimum relative width of the object being recognized. |
A minimum object height. | min_obj_rel_height | float | 0 | [0, 1] | The minimum relative height of the object being recognized. |
A minimum size of undetected objects. | motion_min_scale | float | 2.0 | [1, 4] | The minimum value at which previously unrecognized objects are searched for. |
Time to resume searching for objects. | motion_cooldown | float | 0.5 | [0, 4] | Time required to resume searching for new objects in the same frame area (in seconds). |
A low threshold for an object recognition. | weak_obj_score_threshold | float | 0.8 | [0, 1] | The lowest threshold to estimate an object recognition. If the value is lower than the threshold value, the objects are discarded and aren’t considered for further recognition. |
A minimum threshold value for an object detection. | strong_obj_score_threshold | float | 0.95 | [0, 1] | A minimum threshold value for an abject to be detected. |
An area size for re-detection. | glance_expansion | float | 3.5 | [1, 5] | Defines an area size for re-detection of the tracked objects. Large values may increase tracking of fast-moving objects, but it can also lead to missed object detection due to smaller relative size of objects in the detection area. |
Event processing time. | aggregation_time | float | 0.5 | [0, 10] | The time, measured in seconds, during which the best captured image is selected for object recognition. Larger values of the parameter increase the recognition quality but also increase the time required for an event generation. |
A minimum similarity threshold of the objects. | similarity_threshold | float | 0.25 | [0, 1] | If the similarity level of two captured images is lower than the specified value, two separate events will be registered. Larger values may lead to excess of events, while lower ones can cause missed events. |
A minimum low quality threshold of a frame. | low_quality_threshold | float | 0.3 | [0, 1] | A minimum low-quality threshold value of a frame to be acceptable for recognition. |
A track time of an object. | track_lifetime | float | 2.0 | [0, 10] | A track time of an object that is measured in seconds and doesn’t include new events. |
A similarity threshold of the objects. | tracking_similarity_threshold | float | 0.7 | [0, 1] | The minimum similarity of objects at which an object can be considered a continuation of the track. Lower values can improve stability, but unrelated objects can be grouped into one track. |
A recognized face persistency time. | recognized_face_persistency | float | 10.0 | [0, 600] | Recognized face persistency time. The time, measured in seconds, during which the persistence of the recognized face is active and prevents the duplication of events. |
An event generation interval for a static object. | recognized_face_report_interval | float | 30.0 | (0, 600] | The time, measured in seconds, during which a new event ins’t generated for the recognized face. |
A maximum number of objects under recognition. | tracked_objects_limit | int | 32.0 | [0, 512] | The maximum number of objects under simultaneous recognition. If the limit is exceeded, only the largest objects are saved. |
*In the Acceptable range column, square brackets “[” and/or “]” mean that the threshold values can be used as the default ones; parentheses “(” and/or “)” mean that the threshold values cannot be used as the default values.