Overview#

FaceStream conducts several functions:

Stream reading

Web-cameras, USB and IP-cameras (via RTSP protocol), video files and images can act as data sources.

Stream processing

It searches for faces and bodies in the stream and tracks them until they leave the frame or are blocked.

Liveness check

Liveness check is performed on one or more frames of the track.

Sending face or body bestshots as HTTP-requests onto external service

VisionLabs Software LUNA PLATFORM 5 acts as an external service.

FaceStream workflow depends on the setting of four configurations.

Streams management configuration set in LUNA Streams

Here you can set the settings regarding stream sources such as source type, source address, filtering settings, etc. The settings are set by sending requests with a body in JSON format to the LUNA Streams service. FaceStream takes the settings from LUNA Streams for further processing. A detailed description of how FaceStream works with LUNA Streams is given in the "Interaction of FaceStream with LUNA Streams" section.
FaceStream settings set in LUNA Configurator

Here you can set general FaceStream settings, such as logging, setting up sending images from FaceStream to external services, debugging, etc.
TrackEngine settings set in LUNA Configurator

Here you can set general TrackEngine settings regarding the face or body detection and tracking.
LUNA Streams settings set in LUNA Configurator

Here you can set general settings for the LUNA Streams service, such as logging, database settings, address of the LUNA Licenses service, etc.
FaceEngine settings set in "faceengine.conf" configuration file and transferred during the launch of the FaceStream container.

Here you can set the settings for face recognition. It is recommended to change the parameters in this configuration only in consultation with VisionLabs employees.

The following features are also available when working with FaceStream:

Dynamic creation, editing, and deletion of stream sources via API requests
Real time video streams preview in a browser for the streams with specified parameters
Stream metrics (number of streams, number of errors, number of faces, number of skipped frames, FPS)

FaceStream can be configured to work with either faces or bodies. Simultaneous processing of faces and bodies is not possible.

FaceStream workflow with faces and bodies#

FaceStream can handle both faces and bodies. Each object has its own scheme of operation and its own set of parameters described below.

Simultaneous processing of faces and bodies is not possible, it is necessary to enable processing of either faces or bodies (see the section Parameters "use-face-detector" and "use-body-detector").

The required minimum parameters for working with both objects can be found in the section "Priority parameters list".

FaceStream workflow with faces#

FaceStream application workflow with faces is shown in the image below:

FaceStream receives video from a source (IP or USB camera, web-camera, video file) or images. FaceStream can work with several sources of video streams (the number is set by the license). Sources are set by sending requests with the necessary parameters to the LUNA Streams service;
FaceStream decodes video frames;
The ROI area is cut out from the frame if the "roi" (streams management configuration) parameter is specified;
The received image is scaled to the "scale-result-size" (TrackEngine configuration) size if the "detector-scaling" (TrackEngine configuration) is set;
Faces are detected in the frame;
The face is redetected in the frame instead of detection if the "detector-step" parameter (TrackEngine configuration) is set;
A track is created for each new face in the stream; then it is reinforced with new detections of this face from the subsequent frames. The track is interrupted if the face disappears from the frame. You can set the "skip-frames" parameter (TrackEngine configuration) so the track will not be interrupted immediately, and the system will wait for the face to appear in the area for several frames;
FaceStream filters the frames of low quality and selects bestshots. There are several algorithms for choosing the best detection(s) in the track. See the "Filtering section" (streams management configuration);
If the frame is bestshot, it is added to the collection of bestshots. Depending on the "number_of_bestshots_to_send" (streams management configuration) setting one or several best detections are collected from each track;
Optional. If the "warp" type is set in the "portrait_type" (streams management configuration) parameter, the bestshots are normalized to the LUNA PLATFORM standard, and normalized images are created. Normalized image is better for processing using LUNA PLATFORM;
The bestshots are sent to an external service via HTTP-request. The image may be sent as it is or transformed into the normalized image.

The frequency of images sending is specified in the "sending" (streams management configuration) section.

The sending parameters and external service address are specified in sections "data" (streams management configuration) and "sending" (FaceStream configuration).

FaceStream workflow with bodies#

FaceStream application workflow with bodies is shown in the image below:

FaceStream receives video from a source (IP or USB camera, web-camera, video file) or images. FaceStream can work with several sources of video streams (the number is set by the license). Sources are set by sending requests with the necessary parameters to the LUNA Streams service;
FaceStream decodes video frames;
The received image is scaled to the "scale-result-size" (TrackEngine configuration) size if the "detector-scaling" (TrackEngine configuration) is set;
Bodies are detected in the frame;
The body is redetected in the frame instead of detection if the "detector-step" parameter (TrackEngine configuration) is set;
A track is created for each new body in the stream; then it is reinforced with new detections of this body from the subsequent frames. The track is interrupted if the body disappears from the frame. You can set the "skip-frames" parameter (TrackEngine configuration) so the track will not be interrupted immediately, and the system will wait for the body to appear in the area for several frames;
FaceStream filters low quality frames and selects the bestshots. See "Min-score" (streams management configuration);
If the frame is bestshot, it is added to the collection of bestshots. Depending on the "number_of_bestshots_to_send" (streams management configuration) parameter one or several best detections are collected from each track;
The bestshots are normalized to the LUNA PLATFORM standard, and normalized images are created. Normalized image is better for processing using LUNA PLATFORM;
The bestshots are sent to an external service via HTTP-request. Events can be generated in an external service according to the specified handler (see the description of the event in the LUNA PLATFORM administrator manual). The bestshots are transformed into warps. Along with the bestshots, the coordinates of the human body can be sent if their number is set in the "minimal_body_track_length_to_send" parameter (streams management configuration).

The frequency of images sending is specified in the "sending" (streams management configuration) section.

The sending parameters and external service address are specified in sections "data" (streams management configuration) and "sending" (FaceStream configuration).

Interaction of FaceStream with LUNA Streams#

To work with FaceStream, you should first launch an additional service - LUNA Streams (the default port is 5160). In the request body to the LUNA Streams service, settings for stream management are specified. After sending the request, a stream is created, whose settings are taken by FaceStream for further processing. See the LUNA Streams Open API Specification for request examples.

LUNA Streams has its own user interface designed to work with streams. For more information, see "LUNA Streams user interface".

To use the LUNA Streams service, you should use the LUNA PLATFORM 5 services - LUNA Licenses and LUNA Configurator, as well as PostgreSQL or Oracle and Influx.

The Influx database is needed for the purposes of monitoring the status of LUNA PLATFORM services. If necessary, monitoring can be disabled.

The FaceStream documentation does not describe the use of an Oracle database.

If necessary, you can launch LUNA Streams without LUNA Configurator. This method is not described in the documentation.

FaceStream is licensed using the LUNA PLATFORM 5 key, which contains information about the maximum number of streams that LUNA Streams can process. The license is regulated by the LUNA Licenses service.

See the FaceStream installation manual for detailed information on activating the LUNA Streams license.

The PostgreSQL/Oracle database stores all the data of LUNA Streams.

The general process of interaction between FaceStream and LUNA Streams is presented below:

After sending an HTTP request with the specified parameters to the LUNA Streams (1) service, the presence of a parameter regulating the number of streams for LUNA Streams operation is checked at the LUNA PLATFORM key using the LUNA Licenses (2) service. The number of streams already being processed at the time of the request is also checked using the FaceStream report (7) (see below).

If the key parameter is missing, a license error will be issued.

If at the time of stream creation the maximum number of available streams is not processed yet, the parameters are added to the LUNA Streams database (3) under the unique identifier stream_id. The stream with the parameters gets into the queue (4), where it is in the status "pending" until a special FaceStream worker picks up the stream from the queue for subsequent processing.

You can view the streams in the queue by filtering them in a certain way using the "streams/processing/queue" GET request.

If the maximum amount is already being processed at the time of stream creation, LUNA Streams will not be able to add parameters to the database and a license error will be issued.

If FaceStream is disabled at the time of stream creation, then only the number of streams with the “pending” status that is stipulated by the license can be created. After the FaceStream is launched, the streams created in the queue order will be accepted for processing.

Streams can be created with the status "pause". In this case, they will be added to the database and will wait for a manual status update to "pending".

The queue is implemented in the LUNA Streams service itself and is not external.

Next, FaceStream workers take the parameters of the stream(s) from the queue (5) with the status "pending" and begin processing. In this case, the status of the processed streams is changed to "in_progress" and the stream is removed from the queue.

During processing, data is regularly sent to the main services of LUNA PLATFORM 5 for further processing of frames according to the specified handler_id and for creating events (6), and a report on processing streams in LUNA Streams (7) is regularly sent.

The time of sending reports is fixed and cannot be changed.

If the report says that some stream has been processed, then the FaceStream handler takes the following parameters of the stream with the status "pending" from the LUNA Streams queue (5), and the service changes the status of the stream from "pending" to "in_progress", removing it from the queue. If, for unknown reasons, the report was not transferred, then the streams are re-queued.

Along with the report, additional information about the stream is transmitted, such as sizes and frame rate, bitrate, group of frames (gop), creation time, stream processing start time, last processing error, etc. Preview information is also returned - the URLs of preview and the last frame.

For more detailed description of LUNA Streams stream processing, see "Stream processing pipeline".

Settings for stream management are set using a POST request to the "/streams" resource.

In addition, the following actions are available for a stream:

get existing streams by their "stream_id" with a description of the data of each stream ("get streams" request)
get all information about a stream by its "stream_id" ("get stream" request)
delete existing streams by their "stream_id" ("delete streams" request)
delete stream by its "stream_id" ("remove stream" request)
get the number of streams created ("count streams" request)
update the "description" and "status" fields of a stream by its "stream_id" ("update stream" request)
replace of all stream data with new ones by its "stream_id" ("put stream" request)
get link to the last frame ("get last frame preview" request)
get link to live stream ("get live preview" request)
get streams logs ("get streams logs" request)
delete streams logs ("delete streams logs" request)

A detailed description of requests and example requests can be found in the Open API specification of LUNA Streams service.

Stream distribution in LUNA Streams#

As mentioned earlier, the ability to process multiple streams at the same time is available.

For each stream, its current status is assumed:

pending - stream is waiting for FaceStream worker
in_progress - stream processing is in progress
done - stream processing is completed (relevant for stream transmission types "videofile"/"images" or ["tcp/udp"] with "endless" set to "false")
pause - stream processing is paused by user (not applicable for stream transmission types "videofile" or "images")
restart - stream processing is restarted by server
cancel - stream processing is cancelled by user
failure - stream processing is failed by FaceStream worker
handler_lost - stream processing worker is lost, needs to be passed to another worker (not applicable for stream transmission types "videofile" or "images")
not_found - stream was removed during the processing
deleted - stream was removed intentionally

Statuses "pause" and "cancel" can be specified when updating a stream using the "update stream" request.

Statuses "restart", "handler_lost" are transient. With these statuses, it is impossible to receive a stream, however, the transition through these statuses is logged as usual. The "restart" status can only occur when using the "autorestart" section (see the "Streams automatic restart" section below).

The "not_found" status is internal and will be sent back for feedback if the stream was removed during processing. With this status, it is impossible to receive a stream.

The "deleted" status is virtual. Stream with this status cannot exist, but this status can be seen in the stream logs.

Statuses transition table#

The following table shows statuses that may be received after each listed status.

The "+" symbol means that the status listed in the first row may occur after the status in the first column. An empty field means that there are no cases when the status may occur.

The "-" symbol means that there is no stream in the system (it was not created or it was already deleted).

	-	pending	in_progress	done	restart	pause	cancel	failure	handler_lost
-		+				+
pending	+		+		+	+	+
in_progress	+	+		+	+	+*	+	+	+
done	+	+			+	+
restart		+				+
pause	+	+			+		+
cancel	+	+			+	+
failure	+	+			+	+
handler_lost					+

* not supported for stream transmission types "videofile" or "images"

Stream processing pipeline#

By default, the new stream is created with the "pending" status and immediately enters the processing queue. Stream processing can be postponed by specifying the pause status when creating.

As soon as a free stream worker appears with a request for a pool from the queue, the stream is accepted for processing and it is assigned the "in_progress" status.

After the stream has been processed by the worker, it is assigned to the status "done" in case of success (relevant for stream transmission types "videofile"/"images" or ["tcp/udp"] with "endless" set to "false"), or "failure" if any errors have occurred. However, stream processing status may be downgraded from "in_progress" for the following reasons:

no feedback from stream worker: process will be downgraded by server and record with "handler_lost" status will be added to the stream logs
replacing the stream by user: record with "restart" status will be added to the stream logs

For stream transmission types "tcp" or "udp" with "endless" set to "true", the status cannot change to "done".

During the processing routine, any change in the stream status is logged. Thus, you can restore the stream processing pipeline from the logs.

Streams with "failure" status can be automatically restarted.

Streams automatic restart#

The ability to automatically restart streams is relevant only for streams with a "failure" status. Automatic restart options (restart possibility, maximum number of restart attempts, delay between attempts) are specified by the user for each stream in the "autorestart" section of stream management settings. The parameters and automatic restart status can be received using the "get stream" request.

The automatic restart statuses are listed below:

"disabled" - stream automatic restart is disabled by user ("restart" parameter is disabled)
"enabled" -automatic restart is enabled but is not currently active because the stream is not in the "failure" status
"in_progress" - automatic restart in progress
"failed" - the allowed number of automatic restart attempts was exceeded and none of the attempts were successful
"denied" - automatic restart is allowed by the user, but not possible due to a fatal error* received in the FaceStream report.

* fatal error is considered a Failed to authorize in Luna Platform error.

The process of performing automatic restart of streams:

its status will be automatically changed to "restart" and than, to "pending"
"current_attempt" parameter will be increased by 1
"last_attempt_time" parameter will be actualized

The possibility of autorestart, the maximum number of restart attempts, the delay between attempts are specified by the user for each stream at "autorestart" section.

Stream will be considered successfully restarted if after the specified time period (delay) it has a status other than "failure".

The number of simultaneous processing streams (statuses "pending" and "in_progress") is regulated by the license, but the LUNA Streams database can store an infinite number of streams with a different status, for example, "pause".

Streams grouping#

Streams can be grouped. Grouping is intended to combine streams with multiple cameras into logical groups. For example, you can group streams by territorial characteristic.

A stream can be linked to several groups.

The group is created using the "create group" request. To create a group, you need to specify the required parameters "account_id" and "group_name". If necessary, you can specify a description of the group.

Stream can be linked to a group in two ways:

using the "group_name" or "group_id" parameters during stream creation ("create stream" request).
using the "linker" request. In the request, you should specify the streams IDs and the group to which they need to be linked.

Using the "linker" request you can also unlink streams from a group.

If the stream was linked to a group, then the "get stream" or "get streams" requests will show the group in the "groups" field.

LUNA Streams database description#

The LUNA Streams database general scheme is shown below.

See "Streams management configuration" for a description of the database data.

Recommendations for FaceStream configuration#

This section provides general guidelines for setting up FaceStream.

The names of the configuration, which describes the configured parameters, are mentioned in this section.

Before starting configuration#

You should perform the FaceStream configuration for each camera used separately. FaceStream should work with the stream of the camera, located in the standard operating conditions. The following reasons lead to these requirements:

Frames with different cameras may differ by:
noise level,
frame size,
light,
blurring,
etc.;
FaceStream settings depend on the lighting conditions, therefore, will be different for the cameras placed in a dark room and a light;
FaceStream performance depends on the number of faces or bodies in the frame. Therefore, the settings for the camera, which detects one face every 10 seconds, will be different from the settings for the camera detecting 10 faces per second;
The number of detected faces and bodies and the quality of these detections depend on correct location of the camera. When the camera is at a wrong angle, faces are not detected in frames. Moreover, head angles can also exceed the acceptable degree hence the frame with the detected face could not be used for further processing.
Faces and bodies in the zone of camera view can be partially or completely blocked by some objects. There can be background objects that can prevent the proper functioning of recognition algorithms.

The camera can be positioned so that the lighting or shooting conditions change throughout the day. It is recommended to test FaceStream work under different conditions and choose the best mode, providing reliable FaceStream operation under any conditions.

You can specify the FPS for video processing using the "real_time_mode_fps" parameter.

The video cameras tested with FaceStream are listed in section "Appendix A: Cameras Compatibility".

FaceStream performance configuration#

The mentioned above parameters have the greatest impact on the FaceStream performance.

Reduction of face search area#

Not all the areas of the frame contain faces. Besides, not all the faces in the frame have the required size and quality. For example, the sizes of faces in the background may be too small, and the faces near the edge of the frame may have unacceptable pitch, roll, or yaw angles.

The "roi" parameter (streams management configuration, section "data"), enables you to specify a rectangular area to search for faces.

The specified rectangular area is cut out from the frame and FaceStream performs further processing using this image.

The smaller the search area, the less resources are required for processing each frame.

Correct exploitation of the "roi" parameter significantly improves the performance of FaceStream.

The parameter should be used only when working with faces.

Frame scaling#

Frame scaling before processing can significantly increase the performance of FaceStream. Frame scaling is enabled using the "detector-scaling" parameter.

Frame size suitable for work should be selected using the "scale-result-size" parameter. sets the maximum frame size after scaling the largest side of the frame. If the source frame had a size of 1920x1080 and the value of "scale-result-size" is equal to 640, then FaceStream will process the frame of 640x360 size. You should gradually scale the frame and check whether face or body detection occurs on the frame, to select the optimal "scale-result-size" value. You should set the minimum image size at which all objects in the area of interest are detected.

See an example of using the "scale-result-size" parameter in the appropriate section.

Defining area with movement#

Three parameters are responsible for determining the area with movement, set in the TrackEngine configuration:

frg-subtractor - enable the mode of considering for movements in the frame;
frg-regions-alignment - set alignment for areas with motion;
frg-regions-square-alignment - set equal width and height of the area with movement.

Below are the recommended values for these settings when using CPU and GPU:

	frg-subtractor	frg-regions-alignment	frg-regions-square-alignment
Recommended value when utilizing CPU	1	0	0
Recommended value when utilizing GPU	1	360	0

Batch processing of frames#

Three parameters are responsible for batch processing of frames, set in TrackEngine configuration:

batched-processing - enable batch processing of frames;
min-frames-batch-size - minimal number of frames collected from all the cameras before processing;
max-frames-batch-gather-timeout - time between processing of the batches.

Minimal face size#

You should configure the "minFaceSize" parameter in the FaceEngine configuration file to specify the minimal face size for detection.

You should set the maximum possible face size. The larger the face, the fewer resources are required to perform detections.

Note that the face size will depend on the actual frame size set by the "scale-result-size" parameter (TrackEngine configuration). A face with a size equal to 100 pixels on a 1280x760 frame will have a size equal to 50 pixels on a 640x480 frame.

General configuration information#

Working with track#

A new track is created for each detected face or body. Bestshots are defined and sent for each track.

In general, the track is interrupted when the face can no longer be found in the frame. If a track was interrupted and the same person appears in the frame, a new track is created.

There can be a situation when two faces or bodies interact in a frame (one person behind the other). In this case, the tracks for both persons are interrupted, and new tracks are created.

There can be a situation when a person turns away, or a face or body is temporarily blocked. In this case, you can specify the "skip-frames" parameter (TrackEngine configuration) instead of interrupting the track immediately. The parameter sets the number of frames during which the system will wait for the face to reappear in the area where it disappeared.

When working with the track, it is also useful to use the "detector-step" parameter, which enables you to specify the number of frames on which face redetection will be performed in the specified area before face detection is performed.

Bestshot sending#

The "sending" parameters group (streams management configuration) enables you to set parameters for the bestshot sending. FaceStream sends the received bestshots to LUNA PLATFORM (see "Priority parameters list").

You can send several bestshots for the same face or body to increase the recognition accuracy. You should enable the "number_of_bestshots_to_send" (streams management configuration) parameters in this case.

LUNA PLATFORM enables you to aggregate the bestshots and create a single descriptor of a better quality using them.

If the required number of bestshots was not collected during the specified period or when the track was interrupted the collected bestshots are sent.

The "time_period_of_searching" and "silent_period" parameters (streams management configuration) can be specified in seconds or in frames. Use the "type" parameter to choose the type.

The general options for configuring the "time_period_of_searching" and "silent_period" parameters of the "sending" group from streams management configuration are listed below.

The bestshot is sent after the track is interrupted and the person left the video camera zone of view.

All the frames with the person's face or body are processed and the bestshot is selected.

time_period_of_searching = -1 silent_period = 0

It is required to quickly receive the bestshot and then send bestshots with the specified frequency.

For example, it is required to send a bestshot soon after an intruder entered the shop. The intruder will be identified by the blacklist.

The mode is also used for the demonstration of FaceStream capabilities in real-time.

The bestshot will be sent after the track is interrupted even if the specified period did not exceed.

time_period_of_searching = 3 silent_period = 0

It is required to quickly send the bestshot and then send the bestshot only if the person is in the frame for a long time.

time_period_of_searching = 3 silent_period = 20

It is required to quickly send the bestshot and never send the bestshot from this track again.

time_period_of_searching = 3 silent_period = -1

Frames filtration#

The filtration of face frames is performed by three main criteria (they are all set in the streams management configuration):

Head angles ("detection_yaw_threshold", "detection_pitch_threshold", "detection_roll_threshold").

The "yaw_number" and "yaw_collection_mode" parameters are additionally set for the yaw angle. The parameters reduce the possibility of the error occurrence when the "0" angle is returned instead of a large angle.

Frame quality for further processing ("min_score");
Mouth occlusion ("mouth_occlusion_threshold").

If a frame did not pass at least one of the specified filters, it cannot be selected as a bestshot.

If the "number_of_bestshots_to_send" parameter is set, the frame is added to the array of bestshots to send. If the required number of bestshots to send was already collected, the one with the lowest frame quality score is replaced with the new bestshot if its quality is higher.

The filtration of body frames is performed only by one criterion - "min_score".

Working with ACMS#

Work with ACS is performed only with faces.

Use the "primary_track_policy" settings when working with ACMS. The settings enables you to activate the mode for working with a single face, which has the largest size. It is considered, that the face of interest is close to the camera.

The track of the largest face in the frame becomes primary. Other faces in the frame are detected but they are not processed. Bestshots are not sent for these faces.

As soon as another face reaches a larger size than the face from the primary track, this face track becomes primary and the processing is performed for it.

The mode is enabled using the "use_primary_track_policy" parameter.

The definition of the bestshots is performed only after the size (vertical) of the face reaches the value specified in the "best_shot_min_size" parameter. Frames with smaller faces can't be the bestshots. When the face detection vertical size reached the value set in the "best_shot_proper_size" parameter the bestshot is sent as a bestshot at once.

The "best_shot_min_size" and "best_shot_proper_size" are set depending on the video camera used and its location.

The examples below show configuration of the "sending" group parameters from streams management configuration for working with ACMS.

The turnstile will only open once. To re-open the turnstile you should interrupt the track (move away from the video camera zone of view).

time_period_of_searching = -1 silent_period = 0

The turnstile will open at certain intervals (in this case, every three seconds) if a person stands directly in front of it.

time_period_of_searching = 3 silent_period = 0

If the "use_primary_track_policy" parameter is enabled, the bestshot is never sent when the track is interrupted.

Formats, video compression standards, and protocols#

FaceStream utilizes the FFMPEG library to convert videos and get a stream using various protocols. All the main formats, video compression standards, and protocols that were tested when working with FaceStream are listed in this section.

FFMPEG supports more formats and video compression standards. They are not listed in this section, because they are rarely used when working with FaceStream.

Video formats#

Video formats that are processed using FaceStream:

AVI,
MP4,
MOV,
MKV,
FLV.

Encodings#

Basic video compression standards that FaceStream works with:

MPEG4,
MS MPEG4,
MS MPEG4v2,
MJPEG,
H.264,
H.265.

Protocols#

Basic transport layer protocols used by FaceStream to receive data:

TCP,
UDP.

Basic application layer protocols used by FaceStream to receive data:

HTTP (based on the transport layer protocol TCP)
RTP (based on the transport layer protocol UDP)
RTSP (based on transport layer protocols TCP or UDP)
RTMP (based on the transport layer protocol TCP)
HLS (based on the transport layer protocol TCP)

To use application layer protocols, you must specify the appropriate transport layer protocol in the "type" parameter of the streams management settings.

Memory consumption when running FaceStream#

This section lists the reasons for increasing RAM consumption when running FaceStream.

Each stream increases memory consumption. The amount of the consumed memory depends on the settings set for FaceStream:
the number of Ffmpeg threads in the "ffmpeg_threads_number" parameter (streams management configuration),
image cache size in the "stream_images_buffer_max_size" parameter (FaceStream configuration),
set buffer sizes in the "frames-buffer-size" parameter (TrackEngine configuration).
If the number of threads specified in the "ffmpeg_threads_number" parameter is greater than "1" (streams management configuration), the memory consumption increases significantly. At the same time, the increase in consumption is extremely slow and can be noticed after several hours of operation only.

For RTSP streams, you can set the "ffmpeg_threads_number" parameter to "0" or "1" (streams management configuration). In this case, memory growth is not noticed.

Memory consumption increases after FaceStream starts. Growth occurs within 1-2 hours. This is related to caches filling (see point 1). If no new streams are created and step 2 is not executed, the memory consumption stops growing.
Memory consumption increases when settings in the Debug section are enabled (FaceStream and TrackEngine configurations).

Stream playback interface#

FaceStream has the ability to view the stream in real time. To view the stream, you should enter the following address in the browser bar after the FaceStream starts processing the stream:

http://127.0.0.1:34569/api/1/streams/preview/<stream_id>.

When objects appear in the camera's field of view, FaceStream displays them in a certain way.

Yellow bounding box occurs if a detection fails at least one of the "detection_yaw_threshold", "detection_pitch_threshold" or "detection_roll_threshold" parameters.

Red color bounding box occurs if the detection acceptance score is lower than the value specified in the "min_score" parameter.

Blue color bounding box occurs when an object is detected (redetected) or tracked.

Green bounding box occurs in all other cases when all conditions are met.