Changelogs

Hello, in this part we introduce a new way for us to explain to you what we've improved/changed. Please use these changelogs to monitor how we improve our products!

August 2024 Updates

[4.57.14] - August 12, 2024

Changed

• Make global default face min size from 10px to 48px

• Update bounding box style for FR

• Improve counting logic

• Update draw stylish bounding box on MPA and MVA

• Change face candidate filtering by employing the detection confidence and std_dev

• Update defaut env FR_STDDEV_THRESHOLD=0.7 and FR_PRIMARY_THRESHOLD=0.7

Added

• Add sharpness_score to cutoff blur image when pooling the face candidates and add sharpness score as attribute to contribute for face scoring

• Add default env FR_BLUR_FILTER_WEIGHT=2 and FR_BLUR_FILTER_THRESHOLD=4 for global config. Add fr_blur_filter_threshold for local pipeline config

• Interactive bounding box by updating color of the bounding box into orange if the object cross the line counter (MPA & MVA counting)

• forward softmax config to payload for NFV4D-OVIC

• Add new argument: --disable-warn-visual-api to disable visualization of warning message API to FremisN & Fremis Platform analytic

• Implementing snapshot mode on stream engine by adding postfix capture_mode=snapshot and capture_period=5 on url with separated semicolon ; eg. ;capture_mode=snapshot;capture_period=5

• Expose std_dev on pipeline_data json FR event dump

• Expose env V4_STREAM_OPEN_TIMEOUT(default 8) and V4_STREAM_READ_TIMEOUT(default 5) to control stream timeout in seconds

Fixed

• Tweak False Positive avoidance strategy and expose FR_DEAD_ID_POOL_SIZE(default 4), set this to large number to disable dump loss tracks

• Fix pipeline duplicated, enhancing robustness of main loop terminated

• Fix error create analytic already exist on a stream not rejected.

July 2024 Updates

[4.57.12] - July 23, 2024

Changed

• Update Ubuntu version 22.04, Cuda version 12.x, TensorRT 8.6.x.x

• Using generated prompt given static event_name on other action recognition analytic

• Rename LPR 1 analytic to not be marked as deprecated.

Added

• Add OpenVino 2024.1.0 library

• Enabling Intel Ultra GPU using OpenVino runtime for (Face Recognition) face detection and face extraction model

• New snapshot analytics NFV4D-ARVM for (configurable) action recognition given event_name using Vision Language Model.

• Enabling Intel Ultra GPU using OpenVino runtime for People Analytic and Vehicle Analytic

Fixed

• Fix LPR 1 analytic not dumping event.

• Fix LPR2/LPRS iou_ratio whether plate.size or vehicle.size is 0

• Fix pipeline duplicated, enhancing robustness of main loop terminated

June 2024 Updates

[4.57.11] - June 28, 2024

Changed

• Update default value of --fr-address argument to "127.0.0.1:4005" to simplify installation.

• Optimize resource when stream is disconnected by:

  • Scale up service demand only when stream is active.

  • Automatically scale down service demand when stream is inactive.

Added

• New snapshot analytics NFV4D-FISO for fire and smoke detection using OVOD.

• New snapshot analytics NFV4D-VMFG for people fighting action recognition using Vision Language Model.

• New snapshot analytics NFV4D-VMVD for vandalism event action recognition using Vision Language Model.

• New snapshot analytics NFV4D-VMBG for ATM burglary event action recognition using Vision Language Model.

• New snapshot analytics NFV4D-PPFA for people with firearrm detection using OVOD.

• New cli argument --fremis-coordinator-address to set default address for all fremis platform related analytic.

April 2024 Updates

[4.57.10] - April 28, 2024

Fixed

• Fix reload pipeline on slave by move license data initialization

• Fix supervisor for reloading app on slave

[4.57.9] - April 28, 2024

Fixed

• Fix MVA to not hit snapshot.

[4.57.8] - April 28, 2024

Changed

• Change response pipelines value while GET /streams

[4.57.7] - April 28, 2024

Changed

• Update NFV4-MPA person detector model to use bytetrack

• Update dwelling logic to use object intersection for dwelling metric

• change multi logics dwelling config from "dwelling_treshold_sec" to "dwelling_treshold_alert_sec"

• Rewrite dwelling logic to be more robust

• Disable color detection for counting logic on multi-logics analytics

• Increase value for ignore dump iou threshold on counting logic for person detection.

• Add pipelines with analytic_id and description while GET /streams

• Update warning message visualization and log for external service integration.

Added

• Add config for MPA counting "dump_ignore_iou_threshold": <float> to configure iou threshold for dump objects to be ignored.

Fixed

• Fix supervisor to not create local config file when not exist

• Fixi MPAA missing address in particular logic

• Fixi model input size for NFV4-MPAA and NFV4-MPAF

[4.57.6] - April 23, 2024

Changed

• Optimize package compilation, change target source to private

• Update to use gcovr for test coverage summary

• Metabase is disabled by default in prod image, to enable use env DISABLE_METABASE=0

• Update default value of several cli arguments to simplify installation

  • cors is enabled by default, to disable use --disable-cors

  • --db-host default from host.docker.internal to 127.0.0.1

  • --db-password default from test to nfvisionaire123

  • --db-name default from postgres to nfvisionaire

Added

• Store credential and platform config to database

• Enable to visualize pipeline error message even when pipeline is running

• Add snapshot error message visual on connection error

• New behavior in multi logics dwelling, when config "dwelling_treshold_sec": <int> is set will also dump when object dwelling is more than the value

Fixed

• Fix problem checking capability when adding FR-ISAPI

• Remove redundant build engine to improve loading time

• Fix warning in pipeline and ISAPI parser

[4.57.5] - April 18, 2024

Added

• Expose FR_STDDEV_THRESHOLD to filter low quality FR dump, also add corresponding API params stddev_threshold(default: 0.6)

• Add FR_STATIC_OBJECT_SENSITIVITY to control how likely static object/poster to be dumped(0-10), the bigger number, the more likely static face will be filtered, also add corresponding API params: static_object_sensitivity(integer 0-10)

March 2024 Updates

[4.57.4] - March 25, 2024

Changed

• Update Flood Monitoring better prompt

Added

• Add signature in snapshot analytics to bypass fremis license

Fixed

• Fix MVA and MPA to not hit snapshot

• Fix load file license and config

[4.57.3] - March 25, 2024

Fixed

• Fix potential memory leak when activating ISAPI capture metadata=true

[4.57.2] - March 22, 2024

Fixed

• Fix visual metadata enabled despite capture_metada:false

[4.57.1] - March 21, 2024

Added

• Add mutex on fr_best_event to prevent FP face recognition

• Add multiple search embeddings for consistency test

• Restructure new config ENV and API fr_timeout_dump, fr_sampling_size, and fr_max_dump_count

[4.57.0] - March 15, 2024

Added

• Expose the ENV setting FR_MAX_DUMP_COUNT to limit the number of each face in frame(default: 1000)

• Expose the ENV setting FR_SAMPLING_SIZE to control face grading pooling(default: 10)

• Expose the ENV setting FR_TIMEOUT_DUMP to control interval between same face dumping in millisecond(default: 2500)

Fixed

• Make the face bounding box visualization more stable

[4.56.9] - March 4, 2024

Changed

• Change max resolution to 2k (1440)

Added

• Add NFV4D-FLOD analytic dispacther for flood monitoring analytic with OVIC.

• Add dump interval on face recognition

• ISAPI Interfation base framework

• ISAPI FR Integration and analytics

Fixed

• Fix dump interval variable declaration on face recognition

February 2024 Updates

[4.56.8] - Feb 26, 2024

Changed

• Add L4 & L40 compatibility in multi_logics

[4.56.7] - Feb 24, 2024

Changed

• Moving parse address payload into each logic config (analytic_id=NFV4-MPAA)

Added

• Add NFV4D-PSMO analytic for person smoking detection with OVOD.

• Add NFV4D-PPHO analytic for person using phone detection with OVOD.

[4.56.6] - Feb 16, 2024

Changed

• Add ENV setting, V4_FORCE_FFMPEG=1 for network environment that prone to packet loss(default is 0)

[4.56.5] - Feb 16, 2024

Changed

• Remove dump interval, decrease timeout dump on face recognition

[4.56.4] - Feb 15, 2024

Changed

• Update LPRS plate-vehicle detection model to v1.1

Added

• Add new face recognition addons to people analytic

• Add new analytic 'NFV4-MPAF' under multi-logics

Fixed

• Fix fr less sensitive dump

January 2024 Updates

[4.56.3] - Jan 19, 2024

Changed

• Change current NFV4-MPAA under counting to new analytic NFV4-PCA

• Move NFV4-MPAA from counting to multi-logics

• Jetson decoder gstreamer pipeline using uridecodebin

• Move fr configuration into master configuration

• Swith mode FR non-experimental to FR experimental, vice-versa

Added

• Add new analytic 'NFV4-MVAA' under multi-logics

• Add new analytic id NFV4-LFR for lite face recognition

• Add new face recognition base class

Fixed

• Add initializing std::map before updating map of lpr event

• Fix draw multiple line and roi at MVA and MPA analytics

December 2023 Updates

[4.56.2] - Dec 19, 2023

Changed

• update supervisor to restart v4 immediately after stopped

• rename --enable-fiber-limit to --disable-fiber-limit to reflect its proper behavior

Added

• Push error message if source video resolution exceed the maximum resolution limit

• new env V4_MAX_STREAM_RESOLUTION with default value 1080

• add api-based platform configuration

• save platform config in file

Fixed

• jetson gstreamer plugin not found in devel image

• Handle connection error to snapshot api for NFV4-MPAA analytic

[4.56.1] - Dec 08, 2023

Added

• New visualization filter on mp1streamer given active logic logic=counting, logic=dwelling, logic=density

• New event dump filter given active logic logic=counting, logic=dwelling, logic=density

• Implementing new visualization and new event dump on multi people analytic

• Add new method "GET" with endpoint /pipeline/<node_num>/<stream_id>/<analytic_id> for retrieve pipeline config

[4.56.0] - Dec 04, 2023

Changed

• Break periodic license check if checked_in=false

• In pipeline main loop, throw an error if checked_in=false

• Change behaviour: Implement reload_node() after check-in to avoid throw an error at the first time initiate pipeline instance

Added

• Update credential by adding DELETE /config/checkin

• Push test driver lp metrics to mlflow

• Add new logic for color detection : object_color under logics

• Add new analytics Multi-logic People Analytic (NFV4-MPA)

• Add new analytics Multi-logic Vehicle Analytic (NFV4-MVA)

November 2023 Updates

[4.55.0] - Nov 21, 2023

Fixed

• Fix slave mode check_in didn't follow master check_in state

Changed

• Support for new GPU architecture (lovelace)

• Tune pipeline scaling for Nvidia L4

• Jetson can run as master

• Add monitoring tools (prometheus and grafana) to jetson image

• Upgrade tensorrt 8.5.1 and cuda 11.8 (from nvcr.io/nvidia/tensorrt:22.12-py3)

• Upgrade onnxruntime to 1.15.1

Added

• Add check_in to ONSPAWNPIPELINE rpc to update check_in value in slave mode

[4.54.1] - Nov 07, 2023

Added

• Add new analytic NFV4-MPAA under counting analytic set

Changed

• Minor update and refractor line_counting logics

[4.54.0] - Nov 03, 2023

Changed

• change license NFV4-CELD to NFV4-HD

Fixed

• Fix FR hybrid not dump (on experimental) due to non availability of landmark

• Fix v4 crash when input wrong credential

Added

• New endpoint /config/checkin to enable check in license from API and GET status check in

• Implement check in license from API and Save credential license to file

• Enable check in license by using read from file

• Add status and message to body response of GET /config/checkin request

October 2023 Updates

[4.53.1] - Oct 23, 2023

Changed

• Add new mechanism for Face Recognition Stream.

• Move line counting from Packages to Logics

• Move roi dwelling from Packages to Logics

• Move roi density from Packages to Logics, update dump mechanism

• Unify docker image for gpu and cpu only mode

• CPU only mode can now create pipeline, but not started

• Pipeline that doesn't require service can run on CPU only mode

Fixed

• Fix pipeline FR crash due deleting pipeline

Added

• Implement signature header on create keyspace and recognition by embeddings (FR)

[4.53.0] - Oct 06, 2023

Added

• Add new NFV4D-WLM under snapshot analytic set

• Add service evaluator to GoldenStandard's test driver

• Publish GoldenStandard's test driver result to nodeflux internal MLFlow

September 2023 Updates

[4.52.6] - Sep 29, 2023

Added

• Add FR tracker id-based and between faces similarity matching for reducing same id multiple dumping

[4.52.5] - Sep 29, 2023

Fixed

• Change streaming loop mechanism to accommodate more severe jitter case(avg cams)

• Make stream disconnection more aggressive(5s-8s) on both regular and child mode

[4.52.4] - Sep 28, 2023

Fixed

• Reload node stuck when stream is inactive

[4.52.3] - Sep 25, 2023

Changed

• Update CE pipeline fraction on jetson

• Change mechanism of delete old event partition

Added

• Expose metrics: number of events partition

• Expose metrics: stream_main_connection_status for rtsp connection status

Fixed

• Normalize embedding on dump event

• Fix hybrid stream could not exit main loop while stream rtsp was paused

[4.52.2] - Sep 12, 2023

Added

• Add flag to disable supression on FR using --fr-disable-supression

• Add flag to disable static dump event on FR using --fr-disable-static-dump

[4.52.1] - Sep 8, 2023

Added

• Add confidence filtering for NFV4D-OVOD

• Add confidence to API payload for NFV4D-OVIC

Changed

• Disable dump for NFV4-OVIC when all classification value is false

[4.52.0] - Sep 7, 2023

Added

• Add new NFV4D-OVOD under snapshot analytic set

• Add new NFV4D-OVIC under snapshot analytic set

• Add transfer streams between nodes via HTTP PUT /streams

• Add V5

• Dump GPS location at Hybrid FR analytics with Body Worn

• Add new streampreset=480p

Changed

• CPU only mode as a master node can spawn pipeline on slave node

[4.51.3] - Sep 4, 2023

• Add new streampreset=480p

August 2023 Updates

[4.51.1 ] - August 15, 2023

• Dump image now will be taken from ROI image

[4.51.0 ] - August 14, 2023

• Update model for NFV4-CELD (Crowd Estimation Low Density)

[4.50.2] - August 9, 2023

• Update model for NFV4-VC (Vehicle Counting City View)

July 2023 Updates

[4.49.7] - July 12, 2023

• Update on MP1 Streamer for CPU-Only or Streambox

[4.49.6] - July 10, 2023

• Update model for NFV4-HC (Head Counting)

• Upgrade CUDA on docker from 11.6.1 to 11.6.2

June 2023 Updates

[4.49.0] - June 23, 2023

• Add NFV4-FR filtering based on bounding box, including new parameters: bbox_filter_active, bbox_filter_threshold

• Add new analytic people counting based on head with NFV4-HC

[4.48.1] - June 22, 2023

• Multiple MJPEG server port support

• Expose NFV4-FR parameters frontal_only, matching_iou_threshold to adjust recognition sensitivity

• Add MP1-Streamer support

May 2023 Updates

[4.47.7] - May 7, 2023

• Update model on NFV4-CE to support varies of people density

• Add new stream controller metrics including stream_ws_event_restarted_latency

• Update model for Vehicle Detection analytics

April 2023 Updates

[4.47.6] - April 11, 2023

• Add new FR analytic metrics including face_secondary_rejection_rate, face_secondary_rejection_rate, and fr_candidate_numbers

• Expose Face Recognition sampling time control: sampling_time_ms

March 2023 Updates

[4.47.3] - March 10, 2023

• Configurable parameter for Water Level Monitoring to control inference sampling

[4.45.8] - March 7, 2023

• Configurable parameter for NVF4-LPR2: Bbox Offset to enlarge image dumping

• Add new class to NFV4-PPE: Glasses

February 2023 Updates

[4.43.0] - February 14, 2023

• New analytic NFV4-GOC with purpose of General Object Counting consist with 80 class. (POC Support only)

[4.42.0] - February 13, 2023

• Streams are instantiated as subprocess

• Add stream_child support

[4.41.11] - February 13, 2023

• Add direction detection for NFV4-LPR2

• Add whitelist/blacklist option ROI for NFV4-LPR2

January 2023 Updates

[4.41.5] - January 19, 2023

• Update CE new model with improvement in pipeline to reduce latency by 100ms

[4.41.3] - January 19, 2023

• Solving case in NFV4-FR where face that too close to camera is not dumped

• Solve false positive case in People Counting

[4.41.0] - January 16, 2023

• New url on /mjpeg endpoint with query params ?fps=xx, ?height=xx, and ?jq=xx

• Enable dashboard stream visual setting with parameter height and jpeg_quality

• Set global configuration on jpeg quality image dump each analytic_id with default analytic_id FR jq=80, otherwise jq=60

• Disable std::terminate if pipeline is unresponsive

[4.41.0] - January 4, 2023

• Fix False Negative on NFV4-PPE

December 2022 Updates

[4.40.3] - December 22, 2022

• New model of NFV4-CE and use counting points as visualization

[4.39.0] - December 22, 2022

• New pipeline NFV4-PPE that provides PPE (Personal Protective Equipment) detection

[4.38.0] - December 15, 2022

• New pipeline NFV4-VD that provides vehicle dwelling information

[4.37.1] - December 13, 2022

• New pipeline NFV4-VCAT to use vehicle recognition (enrollment based) in vehicle counting

[4.36.0] - December 9, 2022

• Introduced a new version of CE : Low Density Crowd Estimation (CELD) which using similar principle with people dwelling.

[4.35.0] - December 2, 2022

• Add method pause and resume abstraction

• Pause frame if it isn't invoking get_active_frame()

• Add check_capability on method PUT

November 2022 Updates

[4.34.0] - November 30, 2022

• Add new analytics LPRS - LPR for special case

[4.33.0] - November 28, 2022

• Assign blacklisted area during assign analytic for FR analytic

[4.32.28] - November 23, 2022

• Concatenate camera metrics to v4 metrics

• add log camera routine and print camera logge

[4.32.24] - November 15, 2022

• Embedding information now available on event as vector of float

• Embedding on event can be enabled by adding --fr-embedding-event-enable on runtime argument

[4.32.23] - November 12, 2022

Added

• Add Face Recognition (FR) threshold to control FR dump

October 2022 Updates

[4.32.18] - October 25, 2022

Added

• Update pipeline for adding FR events age in each track_id

[4.32.17] - October 18, 2022

Fixed

• Fix crowd estimation negative estimate

[4.32.16] - October 12, 2022

Added

• Update person detection model to improve performance and accuracy.

[4.32.15] - October 11, 2022

Added

• new analytic Hybrid Face Recognition (NFV4H-FR)

• Add new analytic data to v4bench

• Add synchronous checking capability at create stream by adding key check_capability equal to true

[4.32.14] - October 7, 2022

Added

• Update model for VC to improve our performance and accuracy

[4.32.13] - October 6, 2022

Fixed

• Vehicle Aerial (NFV4-VAS) fix stuck when detection is empty

September 2022 Updates

[4.32.12] - September 29, 2022

Added

• new analytic Vehicle Aerial Surveillance (NFV4-VAS)

[4.32.11] - September 29, 2022

Added

• Integrate ability camera (Face detection code:NFV4-FR-H) through websocket to face recognition pipeline.

• Transform secondary face landmark value.

[4.32.10] - September 27, 2022

Added

• FR interval dumping for each camera assignment using payload, with default value of 0

[4.32.9] - September 21, 2022

Added

• Update VC-HW detection model to use stream batching TensorRT with fp16 enabled

• Update pipeline fraction for VC, PC, PD, FR, LPR2, and WLM for A2 and A30

[4.32.8] - September 14, 2022

Added

• Test driver face recognition

Fixed

• Fixing switch between primary and secondary face detection

August 2022 Updates

[4.32.7] - August 30, 2022

Changed

• Update crowd model to general, improve high crowd counting.

[4.32.6] - August 16, 2022

Changed

• FR add TensorRT stream batching and fp16 implementation for primary FD (not used)

Fixed

• FR fix secondary FD landmark output

[4.32.5] - August 16, 2022

Added

• add blacklist ROI config to LPR2

• add remove bbox inside blacklist ROI

Fixed

• Fix LPR2 for T4 scaling

[4.32.4] - August 10, 2022

Added

• Internal updates

[4.32.3] - August 4, 2022

Added

• Civilian area code for plate regex in LPR2

July 2022 Updates

[4.32.2] - July 28, 2022

Changed

• Internal minor update.

[4.32.1] - July 27, 2022

Changed

• We have optimized our LPRv2 Performance by increasing the number of maximum pipelines from 4 to 9 pipelines.

Fixed

• fix People Dwelling service.

[4.32.0] - July 25, 2022

Added

• We add new analytic : People Dwelling (NFV4-PD)

[4.31.6] - July 15, 2022

Changed

• We changed Person Detection and use a new batching mechanism and enabled FP16 which increased our performance from 10 pipelines to 37 pipelines (increasing 270% performance) and reduced latency from 43 ms to 20ms (reducing 53% latency).

[4.31.5] - July 14, 2022

Added

• Update model secondary face detection input size which increased analytic performance from 9 pipelines to 14 pipelines.

[4.31.4] - July 13, 2022

Added

• We add a new feature to prepare for something big.

[4.31.3] - July 9, 2022

Added

• We optimize TensorRT preprocess. In this part, we tried to implement it on our Vehicle Counting and reduce our latency from 29ms to 11ms for 1 pipeline, which reduce nearly 62% faster than before.

[4.31.2] - July 7, 2022

Added

• HTTP request error handling in uncertainty dumper (for offline deployment).

• Mechanism to not send data for several tries to our cloud when failed once.

[4.31.1] - July 6, 2022

Added

• Add support for fp16 inference.

Changed

• We implement it on our Vehicle Counting. We improve from 6 pipelines to 24 pipelines (increasing 300% performance) based on our T4 GPU and Intel Xeon 24 threads environment.

• Pipeline health check limit was increased to avoid errors when building our inference engine.

[4.31.0] - June 30, 2022

Changed

• We introduce a new mechanism on our Vehicle Counting to improve our performance. We improve our performance from 6 pipelines to 10 pipelines. Tested on our T4 and Intel Xeon 24 threads environment.

Fixed

• We fix NMS and fix overlap not removing overlapped objects