Release Notes: Deep Learning Streamer (DL Streamer) Pipeline Framework Release 2026.2#

Version 2026.2#

Release Date: September 9, 2026

Key 2026.2 highlights:#

  • Heterogeneous multi-sensor batching: gvastreammux was reimplemented on PTS-based cross-stream batching and now multiplexes video and LiDAR streams in a single pipeline.

  • Flexible generative AI deployment: the new gvagenai backend registry with an openai-http backend lets Vision Language Model inference run locally on Intel hardware or be delegated to a remote service without changing the pipeline.

  • Broader standards-based metadata: segmentation, raw tensor (GstAnalyticsTensorMtd) and 3D object detection results are now carried as upstream GstAnalytics metadata, easing migration away from Intel-specific extensions.

  • New analytics capabilities: dwell time and evaluation-point support in gvaanalytics enable time-in-zone and loitering detection use cases, shipped with a ready-to-run sample.

  • Power-aware optimization: the DL Streamer Optimizer can now optimize pipelines against power consumption in addition to throughput.

  • End-to-end 3D sensor pipeline: three new elements (g3dlidarsrc, g3dobjectfuser, g3drender) complete the LiDAR workflow from source through fusion to rendering entirely within DL Streamer.

  • Expanded model coverage: RF-DETR and RF-DETR-SEG converters add modern transformer-based detection and segmentation models.

  • Improved developer onboarding: a new end-to-end tutorial, restructured documentation site, simplified Get Started flow, and per-source model download scripts.

  • Component updates: OpenVINO 2026.2, NPU driver 1.33, and refreshed GPU drivers.

Deep Learning Streamer (DL Streamer) 2026.2 extends the framework beyond classic video analytics into full 3D sensor processing, broadens standards-based metadata coverage, and opens generative AI inference to remote backends.

Metadata standardization continues: segmentation results, raw tensors and 3D object detections are now carried as upstream GstAnalytics metadata, reducing reliance on Intel-specific extensions ahead of their planned deprecation. The gvagenai element gained a pluggable backend registry with an openai-http backend, so Vision Language Model inference can run locally on Intel hardware or be delegated to a remote service without changing the pipeline.

This release completes the 3D sensing pipeline with three new elements (g3dlidarsrc, g3dobjectfuser, g3drender), so a LiDAR workload can now be sourced, parsed, inferenced, fused and rendered entirely within DL Streamer. The stream multiplexer (gvastreammux) was reimplemented on PTS-based cross-stream batching and gained heterogeneous-source support, allowing video and LiDAR to be batched together in one pipeline.

Note: DL Streamer 2026.2 updates to OpenVINO 2026.2 and NPU driver 1.33. Model downloading has migrated to new per-source scripts (HuggingFace, TIMM, Ultralytics); existing automation that calls the previous scripts should be reviewed.

Deep Learning Streamer (DL Streamer) Pipeline Framework is a streaming media analytics framework, based on GStreamer* multimedia framework, for creating complex media analytics pipelines. It ensures pipeline interoperability and provides optimized media, and inference operations using Intel® Distribution of OpenVINO™ Toolkit Inference Engine backend, across Intel® architecture, CPU, discrete GPU, integrated GPU and NPU.

The complete solution leverages:

  • Open source GStreamer* framework for pipeline management

  • GStreamer* plugins for input and output such as media files and real-time streaming from camera or network

  • Video decode and encode plugins, either CPU optimized plugins or GPU-accelerated plugins based on VAAPI

  • Deep Learning models converted from training frameworks TensorFlow*, Caffe* etc.

  • The following elements in the Pipeline Framework repository:

    Element

    Description

    gvaanalytics

    Provides a framework for custom analytics logic on detected objects and metadata.

    gvaattachroi

    Adds user-defined regions of interest to perform inference on, instead of full frame.

    gvaaudiodetect

    Performs audio event detection using AclNet model.

    gvaaudiotranscribe

    Performs audio transcription using OpenVino GenAI Whisper model.

    gvaclassify

    Performs object classification. Accepts the ROI as an input and outputs classification results with the ROI metadata.

    gvadetect

    Performs object detection on a full-frame or region of interest (ROI) using object detection models such as YOLOv4-v11, MobileNet SSD, Faster-RCNN etc. Outputs the ROI for detected objects.

    gvafpscounter

    Measures frames per second across multiple streams in a single process.

    gvafpsthrottle

    Throttles the frame rate of a pipeline to a specified FPS value.

    gvagenai

    Performs inference with Vision Language Models using OpenVINO™ GenAI, accepts video and text prompt as an input, and outputs text description. It can be used to generate text summarization from video.

    gvainference

    Runs deep learning inference on a full-frame or ROI using any model with an RGB or BGR input.

    gvametaaggregate

    Aggregates inference results from multiple pipeline branches.

    gvametaconvert

    Converts the metadata structure to the JSON format.

    gvametapublish

    Publishes the JSON metadata to MQTT or Kafka message brokers or files.

    gvamotiondetect

    Performs lightweight motion detection on NV12 video frames and emits motion regions of interest (ROIs) as analytics metadata.

    gvapython

    Provides a callback to execute user-defined Python functions on every frame. Can be used for metadata conversion, inference post-processing, and other tasks.

    gvarealsense

    Provides integration with Intel RealSense cameras, enabling video and depth stream capture for use in GStreamer pipelines.

    gvastreammux

    Multiplexes multiple input streams into a single pipeline with batch metadata support.

    gvatrack

    Performs object tracking using zero-term, or imageless tracking algorithms. Assigns unique object IDs to the tracked objects.

    gvawatermark

    Overlays the metadata on the video frame to visualize the inference results.

    g3dinference

    Performs deep learning inference on 3D LiDAR sensor data.

    g3dlidarparse

    Parses 3D LiDAR data for use in analytics pipelines.

    g3dlidarsrc

    Source element that feeds 3D LiDAR data directly into analytics pipelines.

    g3dobjectfuser

    Fuses 3D object detections across multiple sensors into a single coherent object set.

    g3drender

    Renders 3D analytics results with configurable camera projection.

    g3dradarprocess

    Processes 3D radar data for use in analytics pipelines.

For the details on supported platforms, please refer to System Requirements. For installing Pipeline Framework with the prebuilt binaries or Docker* or to build the binaries from the open source, refer to Intel® DL Streamer Pipeline Framework installation guide.

New in 2026.2 Release#

Feature

Description

GstAnalytics segmentation metadata

Segmentation results are now published as standard GstAnalytics metadata for consistent downstream consumption and JSON serialization.

Raw tensor metadata (GstAnalyticsTensorMtd)

New support for carrying raw inference tensors as standard analytics metadata, with accompanying reference documentation.

3D object detection metadata

New metadata type for 3D object detections, documented together with the broader 3D sensor and detection metadata model.

Evaluation-point support in gvaanalytics

Custom analytics logic can now be bound to explicit evaluation points, giving precise control over where in the pipeline rules are applied.

Dwell time analytics

New dwell time logic in gvaanalytics enables time-in-zone use cases, delivered together with a loitering detection sample application.

gvagenai backend registry and openai-http backend

The gvagenai element now supports pluggable inference backends, including a new openai-http backend for remote/hosted VLM inference alongside local OpenVINO GenAI.

RF-DETR and RF-DETR-SEG converters

Added converter implementations enabling RF-DETR detection and RF-DETR-SEG segmentation models.

3D LiDAR source element (g3dlidarsrc)

New source element that feeds 3D LiDAR data directly into analytics pipelines, removing the need for external ingestion tooling.

3D object fuser element (g3dobjectfuser)

New element that fuses 3D object detections across multiple sensors into a single coherent object set.

3D render element (g3drender)

New element for rendering 3D analytics results, with a configurable cam-proj camera projection parameter.

Heterogeneous gvastreammux sources

gvastreammux can now multiplex heterogeneous sources, batching video and LiDAR streams together in a single pipeline.

Loitering detection sample

New Python sample demonstrating dwell-time based loitering detection, including a custom watermark plugin.

Coexistence benchmark sample

New benchmark sample comparing DL Streamer and DeepStream on a license plate recognition workload.

New DL Streamer tutorial

New end-to-end tutorial added to the documentation site to accelerate developer onboarding.

New model downloading scripts

Model download reworked into dedicated per-source scripts for HuggingFace, TIMM and Ultralytics, each with its own model list and requirements file.

Skill Scanner CI workflow

New workflow performing automated security and validation scans on the DL Streamer Coding Agent skill.

Coding agent evaluation prompts and benchmark

Added evaluation prompts and a benchmark definition for measuring DL Streamer Coding Agent skill performance.

diff_report CI tool

New CI tool for visual inspection of failing tests, comparing produced JSON output against ground-truth reports.

Latency tracer and gvafpscounter in monolithic build

Latency tracer and gvafpscounter were backported to the monolithic build, making performance instrumentation available in that configuration.

Updated in 2026.2 Release#

Feature

Description

gvastreammux reimplementation

Reimplemented to use presentation timestamps (PTS) for cross-stream batching, improving synchronization correctness across streams.

gvawatermark GPU blurring

Object blurring is now GPU-accelerated, with a stronger blur ratio for better privacy masking.

gvamotiondetect format coverage

Added I420 and YV12 support for system memory, broadening the range of pipelines that can use motion detection.

gvagenai configuration parsing

Reworked configuration parsing and refreshed the associated samples.

Optimizer power-based optimization

Added power-based optimization and reworked streams functionality to operate together with power optimizations.

Optimizer results handling

Reworked result reporting to properly represent fail states, with additional documentation and stability fixes.

g3dinference tuning

Default score threshold changed from 0 to 0.7 and the 3D detection z-coordinate was adjusted for more accurate output.

g3dlidarparse improvements

Added PTS and duration timestamp management, and corrected a log message unit from floats to points.

OpenVINO update

Updated to OpenVINO 2026.2, with OpenVINO GenAI export requirements aligned to the matching release.

NPU driver update

Updated to NPU driver 1.33 and removed the ZE_ENABLE_ALT_DRIVERS workaround.

GPU driver update

Updated GPU drivers, including the Windows GPU and NPU driver versions.

Documentation site restructure

New samples documentation site, simplified Get Started structure, separate table-of-contents sections, and a revamped main README and documentation landing page.

API reference documentation

Added detailed API reference documentation covering inference output and metadata.

gvaanalytics documentation and discoverability

Updated gvaanalytics documentation and added the gvaanalytics sample to the top-level Python samples README.

Python samples testability

Selected Python samples updated with additional inputs and pinned model versions, with the CI workflow adjusted accordingly.

Smart NVR sample

Updated to comply with DL Streamer Coding Agent guidelines.

Depth estimation sample

Applied changes and fixes to the depth estimation sample.

open_close_valve sample

Updated the model used by the open_close_valve Python sample.

Windows build experience

Enhanced the Windows build script with winget repair and improved Python detection, and added a warning about DLL shadowing.

Dependency updates

Updated pip to 26.1.2, pillow to 12.3.0, onnx to 1.22.0, setuptools, and the GitHub Actions dependency group.

Test execution

Sped up unit tests to avoid timeouts on slow CI machines, standardized test environment setup via setup_dls_env.sh, and adjusted the g3dinference PointPillars test to run with score-threshold=0.

Fixed in 2026.2#

#

Description

1

Fixed a deadlock in gvastreammux state change handling.

2

Fixed an issue with resource management causing leaks under sustained operation.

3

Resolved CodeQL findings raised by static analysis.

4

Fixed Coverity missing lock warnings for thread safety.

5

Fixed incorrect dimension ordering passed into cv::Size in image processing.

6

Fixed HuggingFace model conversion failures.

7

Fixed the queue element destination pad name check.

8

Forced the RAW converter for gvainference regardless of model metadata, ensuring raw output is produced as expected.

9

Fixed failing Windows CI runs.

10

Fixed HTTP Error 403 (Forbidden) when retrieving assets in Python samples.

11

Fixed Python requirements and the shared_utils path in Smart NVR and Face Detection samples.

12

Corrected the setup instructions in the vlm_alerts sample README.

13

Fixed the download paths for the pallet defect detection model.

14

Corrected the SHA in the HuggingFace download README and improved error handling.

15

Correctly handle exported model files and CLI exit codes in model download scripts.

16

Fixed template path resolution in add_vs_version_resource and dropped a redundant project() call in a subdirectory CMakeLists.

17

Added process checking to prevent installation conflicts.

18

Removed references to local GStreamer binaries to avoid conflicts with DL Streamer-installed binaries.

19

Preserved environment variables such as proxy settings when adding the Intel GPU PPA repository.

20

Fixed broken references in Supported Models and other documentation links.

21

Fixed images not rendering in HTML for the Tutorial and Samples docs, and replaced the overview pipeline image with the DL Streamer diagram.

22

Added a missing return value.

23

Fixed issues in hello_dlstreamer.sh.

24

Resolved issues with the Docker image size checker.

Known Issues - 2026.2#

Issue

Description

Preview Architecture 2.0 samples

Preview Architecture 2.0 samples have known issues with inference results and should not be used as a correctness reference.

Model download script migration

Model downloading moved to new per-source scripts. Automation built on the previous scripts may need to be updated; the earlier scripts were restored for compatibility but are considered legacy.

Release Notes: Deep Learning Streamer (DL Streamer) Pipeline Framework Release 2026.1#

Version 2026.1#

Key 2026.1 highlights:#

  • DL Streamer Coding Agent: AI-assisted pipeline builder that translates natural-language descriptions into working DL Streamer Python apps or GStreamer command lines

  • GstAnalytics metadata migration: full-frame GstAnalytics support, keypoints switched to GstAnalytics API, migration to upstream GStreamer 1.28+ API

  • New elements: gvastreammux, gvaanalytics, g3dinference

  • New model support: YOLO classification, PaddleOCRv5, GETI semantic segmentation

  • New D3D11 backend for inference elements for Windows GPU processing

  • PyTorch Image Models (TIMM) model import script

  • Windows installer and expanded Windows sample coverage

  • gvawatermark enhancements: custom full-frame text, dynamic Gaussian blur sized to ROI, metadata-controlled drawing

  • Inference enhancements: CPU thread affinity parameter, skip-raw-tensors property, DepthConverter

  • Optimizer enhancements: time reduction, pause/resume, partial results, file output, cross-stream batch/nireq grouping, detection-based validation

  • Deep SORT tracker refactoring with enhanced detection and track structures

  • New Python samples: gvaanalytics, inference performance optimizations, custom frame selection for VLM queries

  • Dynamic ONVIF camera discovery with DLS pipeline startup

  • DL Streamer ONVIF package as installable Python wheel

  • Component updates: OpenVINO 2026.1, GStreamer 1.28.2, OpenCV 4.13.0, NPU driver 1.32.1, Ubuntu 24.04.4 kernel 6.17

  • Introduced support for Intel® Core™ Processors - series 3 (Wildcat Lake)

  • CI: self-hosted Kubernetes infrastructure, configurable tests, Fedora 41 image building

Deep Learning Streamer (DL Streamer) Pipeline Framework is a streaming media analytics framework, based on GStreamer* multimedia framework, for creating complex media analytics pipelines. It ensures pipeline interoperability and provides optimized media, and inference operations using Intel® Distribution of OpenVINO™ Toolkit Inference Engine backend, across Intel® architecture, CPU, discrete GPU, integrated GPU and NPU.

The complete solution leverages:

  • Open source GStreamer* framework for pipeline management

  • GStreamer* plugins for input and output such as media files and real-time streaming from camera or network

  • Video decode and encode plugins, either CPU optimized plugins or GPU-accelerated plugins based on VAAPI

  • Deep Learning models converted from training frameworks TensorFlow*, Caffe* etc.

  • The following elements in the Pipeline Framework repository:

    Element

    Description

    gvaanalytics

    Provides a framework for custom analytics logic on detected objects and metadata.

    gvaattachroi

    Adds user-defined regions of interest to perform inference on, instead of full frame.

    gvaaudiodetect

    Performs audio event detection using AclNet model.

    gvaaudiotranscribe

    Performs audio transcription using OpenVino GenAI Whisper model.

    gvaclassify

    Performs object classification. Accepts the ROI as an input and outputs classification results with the ROI metadata.

    gvadetect

    Performs object detection on a full-frame or region of interest (ROI) using object detection models such as YOLOv4-v11, MobileNet SSD, Faster-RCNN etc. Outputs the ROI for detected objects.

    gvafpscounter

    Measures frames per second across multiple streams in a single process.

    gvafpsthrottle

    Throttles the frame rate of a pipeline to a specified FPS value.

    gvagenai

    Performs inference with Vision Language Models using OpenVINO™ GenAI, accepts video and text prompt as an input, and outputs text description. It can be used to generate text summarization from video.

    gvainference

    Runs deep learning inference on a full-frame or ROI using any model with an RGB or BGR input.

    gvametaaggregate

    Aggregates inference results from multiple pipeline branches

    gvametaconvert

    Converts the metadata structure to the JSON format.

    gvametapublish

    Publishes the JSON metadata to MQTT or Kafka message brokers or files.

    gvamotiondetect

    Performs lightweight motion detection on NV12 video frames and emits motion regions of interest (ROIs) as analytics metadata.

    gvapython

    Provides a callback to execute user-defined Python functions on every frame. Can be used for metadata conversion, inference post-processing, and other tasks.

    gvarealsense

    Provides integration with Intel RealSense cameras, enabling video and depth stream capture for use in GStreamer pipelines.

    gvastreammux

    Multiplexes multiple input streams into a single pipeline with batch metadata support.

    gvatrack

    Performs object tracking using zero-term, or imageless tracking algorithms. Assigns unique object IDs to the tracked objects.

    gvawatermark

    Overlays the metadata on the video frame to visualize the inference results.

    g3dinference

    Performs deep learning inference on 3D LiDAR sensor data.

    g3dlidarparse

    Parses 3D LiDAR data for use in analytics pipelines.

    g3dradarprocess

    Processes 3D radar data for use in analytics pipelines.

For the details on supported platforms, please refer to System Requirements.

For installing Pipeline Framework with the prebuilt binaries or Docker* or to build the binaries from the open source, refer to Intel® DL Streamer Pipeline Framework installation guide.

New in 2026.1 Release#

Feature

Description

DL Streamer Coding Agent

AI-powered coding assistant that translates natural-language pipeline descriptions into working DL Streamer applications (Python or GStreamer CLI). Ships with a structured skill comprising a requirements questionnaire, model preparation guide, pipeline construction rules, design patterns, sample index, and debugging hints. Seven example prompts included: pose estimation, people detection & tracking, VLM safety compliance checks, event-based smart NVR, license-plate recognition, multi-stream compose, and DeepStream-to-DLStreamer Python conversion.

Upstream GstAnalytics API migration

Removed custom gstanalyticsgroupmtd and gstanalyticskeypointmtd implementations, migrating to upstream GStreamer 1.28+ API.

Full-frame GstAnalytics support

Added full-frame GstAnalytics support so inference results without a parent object detection are properly stored in GstAnalyticsRelationMeta and serialized to JSON. Includes frame-level classification and keypoint conversion.

Keypoints GstAnalytics metadata

Switched keypoints to GstAnalytics metadata API for standardized metadata handling.

Analytics element (gvaanalytics)

New element providing a framework for custom analytics logic on detected objects and metadata.

Stream multiplexer element (gvastreammux)

New element for multiplexing multiple input streams into a single pipeline with GstAnalyticsBatchMeta support.

3D inference element (g3dinference)

New element for performing deep learning inference on 3D LiDAR sensor data.

YOLO classification support

Added YOLO classification model support for image classification tasks.

PaddleOCRv5 model support

Added support for PaddleOCRv5 models for character recognition.

GETI semantic segmentation support

Added support for GETI semantic segmentation models.

CPU thread affinity parameter

Added affinity CPU thread affinity parameter to gvadetect/gvaclassify for performance tuning.

Skip raw tensors property

Added skip-raw-tensors property to gvaclassify and related components to reduce metadata overhead.

DepthConverter and pre-processing

Added DepthConverter and enhanced pre-processing capabilities for depth data.

New D3D11 backend for inference elements

New D3D11 GPU processing backend for inference elements including model sharing support.

Windows installer

Added Windows installer for simplified DL Streamer deployment.

Windows samples expansion

Added new Windows samples including YOLO, metapublish Kafka/MQTT.

Dynamic ONVIF camera discovery sample

New sample for dynamic ONVIF camera discovery with DLS pipeline startup.

gvaanalytics sample

New sample demonstrating the gvaanalytics element usage.

Inference performance optimization sample

New sample app demonstrating inference performance optimizations DLS can provide.

Custom frame selection for VLM sample

New sample pipeline with custom frame selection logic for VLM queries.

DL Streamer ONVIF Python wheel

Introduced dlstreamer.onvif package as an installable Python wheel (whl file).

PyTorch Image Models (TIMM) import

Added PyTorch Image Models (TIMM) model import script for model onboarding.

Python environment setup script

Added Python environment setup script for streamlined development setup.

Updated in 2026.1 Release#

Feature

Description

gvawatermark enhancements

Custom full-frame text support, dynamic Gaussian blur kernel sized to ROI, metadata-controlled watermark drawing, explicit enable-blur flag documentation.

DLS Optimizer enhancements

Time reduction, ability to pause/resume progress, retrieve partial results, output results to file, cross-stream grouping for batches and nireqs, pipeline validation based on counted detections.

Deep SORT tracker refactoring

Refactored Deep SORT Tracker with enhanced detection and track structures for improved tracking accuracy.

metaaggregate improvements

Fixed id=0 hash lookup with g_hash_table_lookup_extended, fixed shared group member duplication, and added generic semantic_tag copy for all mtd types.

gvaattachroi GstAnalytics dual-write

Full-frame keypoints attached by gvaattachroi now also write GstAnalytics metadata for visibility to gvawatermark.

jsonconverter enhancements

Added frame-level keypoint group conversion, frame-level classification via semantic_tag keys, and per-keypoint semantic_tag to JSON output.

ROI label handling

Enhanced ROI label handling: append label regardless of confidence and show confidence only if valid.

YOLOv10/YOLOv26 converters

Updated YOLOv10 and YOLOv26 converters to enforce NMS for improved detection results.

OpenVINO update

Update to OpenVINO 2026.1.

GStreamer update

Update to GStreamer 1.28.2 with GStreamer 1.28 build support. Removed GStreamer 1.26 support.

OpenCV update

Update to OpenCV 4.13.0.

NPU driver update

Update to NPU driver version 1.32.1.

GPU driver update

Updated GPU drivers to latest version.

Ubuntu update

Bump Ubuntu to 24.04.4 and kernel 6.17.

LidarMeta refactoring

Refactored LidarMeta implementation from C++ to C for improved portability.

Model downloading scripts

Improved Ultralytics and HuggingFace model downloading scripts with better error handling. Windows YOLO script now supports YOLOv26.

gvapython samples conversion

Converted gvapython samples to standalone Python GStreamer elements.

gvametaconvert documentation

Added missing documentation about add-rtp-timestamp property.

gvametapublish Windows support

Enabled gvametapublishkafka and gvametapublishmqtt on Windows.

CI infrastructure

Switched to self-hosted Kubernetes infrastructure, enabled configurable tests in Makefile, added Fedora 41 image building, modified PR workflow to save CI resources.

Windows build improvements

Build gstanalytics library from source, PDB file generation, fixed environment variable paths.

Python samples overview

Updated Python samples overview and READMEs.

Tests

Expanded optimizer unit tests, added GStreamer check framework unit tests for gvawatermark, run optimizer tests always. Added 41 unit tests for metaaggregate copy functions, extended tensor_convert_test with full-frame classification and keypoint round-trip tests.

Fixed in 2026.1#

#

Description

1

Fixed downloading of pallet defect detection model.

2

Fixed codeQL scan issues.

3

Fixed Coverity data race warnings in gvastreammux/demux.

4

Fixed GStreamer cache cleaning by prepending directories to PATH.

5

Fixed draw_face_attributes sample failing to build on Ubuntu 24 deb install due to missing OpenCV paths.

6

Fixed instant EOS and tracking metadata copy failure in multi-branch pipelines.

7

Fixed detections with 0 height/width causing issues.

8

Fixed DLSPS building errors.

9

Fixed path for downloading Windows assets.

10

Fixed Windows set environment path issue.

11

Fixed GIR to typelib compilation for Windows.

12

Fixed WinGet PowerShell module crash.

13

Fixed mapping of VA memory in gvawatermark.

14

Fixed incorrect default devices being used in optimizer.

15

Fixed images not showing in docs website.

16

Fixed analytics metadata documentation.

17

Fixed linking dlstreamer to home directory.

18

Fixed triggering CI when pushed to main.

19

Fixed documentation for DL Streamer and DeepStream coexistence.

20

Fixed queue name check and removed mutex unlock (reverted).

21

Fixed yamllint issues.

22

Removed deprecated ffmpeg_openvino and decode_resize_inference samples.

Known Issues - 2026.1#

Issue

Description

Preview Architecture 2.0 Samples

Preview Arch 2.0 samples have known issues with inference results.

Legacy Features and Deprecation Timeline - 2026.1#

List of the features and components to be deprecated in the future.

Feature

Target

Replacement

OMZ models

end of Q3’2026

Use HuggingFace, Ultralytics, TIMM

Model-proc-file

end of Q3’2026

ModelAPI (common with Geti)

gvapython element

end of Q4’2026

usage of regular GStreamer Python bindings

GstVideoRegionOfInterest meta + Intel extensions

end of Q4’2026

usage of GstAnalyticsMtd

Architecture 2.0 elements

end of Q4’2026

Preserve essential components and ideas

Alchemist GPU support

end of Q3’2026

Battlemage discrete GPUs

Tiger Lake support

end of Q4’2026

new supported units: Arrow Lake and Panther Lake

Ubuntu 22.04

end of Q4’2026

newer Ubuntu versions, including Ubuntu 26

Release Notes: Deep Learning Streamer (DL Streamer) Pipeline Framework Release 2026.0#

Version 2026.0#

Key highlights:#

  • New elements: gvafpsthrottle, g3dradarprocess, g3dlidarparse

  • New model support: YOLOv26, YOLO-E, RT-DETR, HuggingFace ViT

  • Streamlined integration with Ultralytics and HuggingFace model hubs

  • GstAnalytics metadata support: DL Streamer supports GstAnalytics metadata for object detection, classification, tracking and adds custom GstAnalytics extension for keypoints

  • gvawatermark overhaul: object bluring, text backgrounds, label filtering, extra fonts, thickness/color options, FPS overlay

  • Inference enhancements: batch timeout, OpenCV tensor compression for all devices

  • Windows platform: GPU inference via D3D11, gvapython support, CI integration, build/setup improvements

  • New Python samples: VLM Alerts, Smart NVR, ONVIF Discovery, face detection/age classification, open-vocabulary detection, RealSense, DL Streamer + DeepStream

  • Optimizer: multi-stream optimization, cross-stream batching, device selection, refactored with tests

  • Component updates: OpenVINO 2026.0.0, NPU driver 1.30, RealSense SDK 2.57.5

  • Library consolidation: merged gvawatermark3d, gvadeskew, gvamotiondetect, gvagenai into gstvideoanalytics

  • CI: Zizmor security scanning, Windows CI, Docker image size checks

Deep Learning Streamer (DL Streamer) Pipeline Framework is a streaming media analytics framework, based on GStreamer* multimedia framework, for creating complex media analytics pipelines. It ensures pipeline interoperability and provides optimized media, and inference operations using Intel® Distribution of OpenVINO™ Toolkit Inference Engine backend, across Intel® architecture, CPU, discrete GPU, integrated GPU and NPU. The complete solution leverages:

  • Open source GStreamer* framework for pipeline management

  • GStreamer* plugins for input and output such as media files and real-time streaming from camera or network

  • Video decode and encode plugins, either CPU optimized plugins or GPU-accelerated plugins based on VAAPI

  • Deep Learning models converted from training frameworks TensorFlow*, Caffe* etc.

  • The following elements in the Pipeline Framework repository:

    Element

    Description

    gvaattachroi

    Adds user-defined regions of interest to perform inference on, instead of full frame.

    gvaaudiodetect

    Performs audio event detection using AclNet model.

    gvaaudiotranscribe

    Performs audio transcription using OpenVino GenAI Whisper model.

    gvaclassify

    Performs object classification. Accepts the ROI as an input and outputs classification results with the ROI metadata.

    gvadetect

    Performs object detection on a full-frame or region of interest (ROI) using object detection models such as YOLOv4-v11, MobileNet SSD, Faster-RCNN etc. Outputs the ROI for detected objects.

    gvafpscounter

    Measures frames per second across multiple streams in a single process.

    gvafpsthrottle

    Throttles the frame rate of a pipeline to a specified FPS value.

    gvagenai

    Performs inference with Vision Language Models using OpenVINO™ GenAI, accepts video and text prompt as an input, and outputs text description. It can be used to generate text summarization from video.

    gvainference

    Runs deep learning inference on a full-frame or ROI using any model with an RGB or BGR input.

    gvametaaggregate

    Aggregates inference results from multiple pipeline branches

    gvametaconvert

    Converts the metadata structure to the JSON format.

    gvametapublish

    Publishes the JSON metadata to MQTT or Kafka message brokers or files.

    gvamotiondetect

    Performs lightweight motion detection on NV12 video frames and emits motion regions of interest (ROIs) as analytics metadata.

    gvapython

    Provides a callback to execute user-defined Python functions on every frame. Can be used for metadata conversion, inference post-processing, and other tasks.

    gvarealsense

    Provides integration with Intel RealSense cameras, enabling video and depth stream capture for use in GStreamer pipelines.

    gvatrack

    Performs object tracking using zero-term, or imageless tracking algorithms. Assigns unique object IDs to the tracked objects.

    gvawatermark

    Overlays the metadata on the video frame to visualize the inference results.

    g3dradarprocess

    Processes 3D radar data for use in analytics pipelines.

    g3dlidarparse

    Parses 3D lidar data for use in analytics pipelines.

For the details on supported platforms, please refer to System Requirements. For installing Pipeline Framework with the prebuilt binaries or Docker* or to build the binaries from the open source, refer to DL Streamer Pipeline Framework installation guide.

New in this Release

Title

High-level description

3D elements (g3dradarprocess, g3dlidarparse)

New 3D plugin support with g3dradarprocess element for radar data processing and g3dlidarparse element for lidar data parsing, enabling 3D analytics pipelines.

FPS throttle element (gvafpsthrottle)

New element to throttle the frame rate of a pipeline to a specified FPS value.

YOLOv26 model support

Added converters and post-processing for YOLOv26 models, including oriented bounding box (OBB) support and INT8 GPU inference. Added YOLOv26 to supported models in samples.

RT-DETR model support

Added RT-DETR support implementation for real-time detection transformer models.

HuggingFace ViT classifier support

Added HuggingFace Vision Transformer (ViT) classifier config parser for inference.

Batch timeout for inference elements

Added batch-timeout parameter to inference elements, allowing control over batching wait time.

VLM Alerts sample

New Python sample for VLM-based alerts with displaying results on produced video.

Smart NVR sample

New Python sample for Smart NVR with added custom analytics logic (gvaAnalytics) and custom storage (gvaRecorder) elements.

ONVIF Camera Discovery sample

New Python sample demonstrating ONVIF camera discovery and DL Streamer pipeline launcher.

Face detection & age classification sample

New Python sample for face detection and age classification using HuggingFace models.

Open-vocabulary object detection sample

New Python sample with open-vocabulary prompt for object detection.

DL Streamer + DeepStream coexistence sample

New sample demonstrating DL Streamer and DeepStream working in one system.

Motion detect sample (Windows)

New sample demonstrating DL Streamer gvamotiondetect functionality

RealSense element usage sample

New sample demonstrating gvarealsense element usage.

Updated in this Release

Title

High-level description

gvawatermark enhancements

Major enhancements to the gvawatermark element: display configuration options (thickness, color index), text background support, inclusive/exclusive label filtering, additional font support, average FPS info overlay, and visual documentation.

DLS Optimizer enhancements

Optimizer refactored with multi-stream optimization, cross-stream batching, improved FPS reporting, and device selection improvements.

gvametaconvert enhancements

Added reference NTP timestamp from RTCP sender meta extraction to gvametaconvert element

ROI object construction enhacement

For existing GstAnalyticsODMtd only, creates GstVideoRegionOfInterestMeta until full GstAnalytics migration

Latency tracer multi-source/sink support

Extended latency_tracer to support multiple sources and multiple sinks.

Detection anomaly converter

Refactored and enhanced anomaly logic in DetectionAnomalyConverter.

FP32 precision in BoxesLabelsConverter

Added FP32 precision support in BoxesLabelsConverter label parsing.

Bounding box validation

Added extra validation of bounding boxes to improve robustness.

OpenCV tensor compression for all devices

Use OpenCV tensor compression for all inference devices, yielding best performance across CPU/GPU/NPU.

Model API refactoring

Moved Model API parser to separate files; added conversion from Ultralytics and HuggingFace metadata to Model API.

Python samples overview

Added overview section for Python samples; updated READMEs.

Tests

Expanded coverage of functional and unit tests.

Windows: GPU inference with D3D11

Added support for GPU inference on Windows using D3D11.

Windows: gvapython support

Added Windows support for gvapython element and gstgva Python bindings.

Windows: enhanced build & setup

Enhanced Windows build/setup scripts, added remove script, Visual C++ runtime handling, and JSON output for Windows samples.

Windows: CI integration

Enabled Windows tests in GitHub Actions workflow, model downloads on Windows.

Library consolidation

Merged gvawatermark3d, gvadeskew, gvamotiondetect, and gvagenai into the gstvideoanalytics library.

OpenVINO update

Update to OpenVINO 2026.0.0.

NPU driver update

Update to NPU driver version 1.30.

RealSense update

Update to Intel RealSense SDK 2.57.5.

Model download script improvements

Simplified YOLO model download script, enhanced INT8 quantization, refactored YOLOv8+ export/quantize, added model validation.

CI: Zizmor security scanning

Added Zizmor GitHub Actions security scanner.

Fixed

#

Issue Description

1

Fixed YOLO26 model inference on GPU FP16/FP32.

2

Fixed threshold parameter in gvadetect not working with PDD model.

3

Fixed yolov8-seg inference result different from OpenVINO.

4

Fixed gvapython failing to read yolo-pose keypoint metadata.

5

Fixed NV12 frame data in Python by removing padding correctly.

6

Fixed watermark default text background behaviour.

7

Fixed check for pad_value in model XML file.

8

Fixed yolo_v10.cpp compile error on Windows.

9

Fixed DLL output paths on Windows.

10

Fixed compilation warnings on Windows.

11

Fixed timestamp on VS 2026.

12

Fixed GStreamer downloader by adding UserAgent.

13

Fixed libva path setup in setup_dls_env.ps1

14

Removed libva dependency for monolithic elements on Windows.

15

Fixed latency tracker for smart intersection pipelines.

16

Fixed environment variable paths in Ubuntu install guide.

17

Fixed directory already exists error during build.

18

Removed duplicate gvametapublish element register.

19

Reverted RTP timestamp feature due to issues.

20

Fixed download public models script - versions of NumPy, Onnx, and Seaborn.

21

Fixed missing context in Build Docker instruction.

22

Fixed formatting in installation guide and developer guide documentation.

Known Issues

Issue

Issue Description

Preview Architecture 2.0 Samples

Preview Arch 2.0 samples have known issues with inference results.

Legacy Features and Deprecation Timeline#

List of the features and components to be deprecated in the future.

Feature

End of Support Date

Replacement strategy

OMZ models

end of Q3’2026

Use HuggingFace, Ultralytics, TIMM

Model-proc-file

end of Q3’2026

ModelAPI (common with Geti)

WSL support

end of Q3’2026

native Windows support

GstVideoRegionOfInterest meta + Intel extensions

end of Q4’2026

usage of GstAnalyticsMtd

Architecture 2.0 elements

end of Q4’2026

Preserve essential components and ideas

FFMpeg integration samples

end of Q2’2026

no replacement