Available Sample Apps

This page indexes all DL Streamer samples by use case. See Using Sample Apps for installation, model download, and run instructions before trying any of the samples below.

Each entry below is a mini “card”: name, a one-line preview thumbnail (when available), what it demonstrates, the elements/models it uses, and its language — CLI (gst-launch command line), Python, or C++.

---

Browse by category

Object detection, classification & segmentation (9)	Object tracking & analytics (3)
Vision-Language Models (VLM) & GenAI (5)	Audio analytics (2)
3D: LiDAR & radar (5)	Cameras & input sources (4)
Metadata: publishing, access & visualization (8)	Customization & extensibility (8)
Performance & benchmarking (2)	Interoperability (2)
Auto-generated reference applications (8)

Tip

Use your browser’s find-in-page (Ctrl+F / Cmd+F) to search across all samples by name, element (e.g. gvadetect), or model (e.g. yolo11n).

---

Object detection, classification & segmentation

Preview	Sample
	Detection with YOLO CLI Object detection and classification with publicly available YOLO models. Key elements: gvadetect, gvaclassify Models: yolox_s (default; many YOLO variants)
	Face Detection and Classification CLI Detect faces and estimate age, gender, emotions and facial landmarks. Key elements: gvadetect, gvaclassify Models: centerface, dima806_facial_age_image_detection, dima806_fairface_gender_image_detection, dima806_face_emotions_image_detection
	Instance Segmentation CLI Instance segmentation with YOLO-seg models visualized by gvawatermark. Key elements: gvadetect, gvawatermark Models: yolo26s-seg (default; also yolo11s-seg)
	Human Pose Estimation CLI Full-frame human pose estimation. Key elements: gvaclassify Models: yolo26s-pose
	Depth Estimation CLI YOLO11n detection followed by Depth Anything V2 depth estimation on detected regions. Key elements: gvadetect, gvainference Models: yolo11n, Depth-Anything-V2-Small-hf
	License Plate Recognition CLI YOLO detector combined with an optical character recognition model. Key elements: gvadetect, gvainference Models: yolov8 license-plate detector, PP-OCRv4
	Prompt-based Object Detection Python Search a video for user-defined objects using an open-vocabulary model (YOLOE). Key elements: gvadetect Models: yoloe-26s-seg (text-prompt, class baked in at export)
	Deployment of Geti™ models CLI Deploy Geti™-trained models for detection, anomaly detection and classification. Key elements: gvadetect, gvaclassify Models: Geti™-trained (Padim / STFPM / UFlow)
	Motion Detect CLI Run detection only over motion ROIs (GPU and CPU paths). Key elements: gvamotiondetect, gvadetect Models: yolov8n

Object tracking & analytics

Preview	Sample
	Vehicle and Pedestrian Tracking CLI Object tracking across frames. Key elements: gvatrack, gvadetect, gvaclassify Models: yolo26s, dima806_vehicle_10_types_image_detection
	Vehicle Counter with gvaanalytics Tripwires Python Count vehicles crossing a virtual line in both directions using tripwires. Key elements: gvaanalytics, gvatrack Models: yolo11n
	Smart NVR for Lane Hogging Detection Python Build an NVR with custom analytics and video storage to detect lane-hogging events. Key elements: gvaanalytics_py, gvarecorder_py Models: rtdetr_v2_r50vd (RT-DETRv2)

Vision-Language Models (VLM) & GenAI

Preview	Sample
	Using VLM Models with gvagenai CLI Video summarization with MiniCPM-V. Key elements: gvagenai Models: MiniCPM-V, Phi-4-multimodal-instruct or Gemma-3
	VLM Alerts Python Edge alerting pipeline that generates structured JSON alerts per frame with annotated video. Key elements: gvagenai Models: Configurable VLM (e.g. Qwen2.5-VL, InternVL)
	VLM-assisted Self Checkout Python Combine CV object detection with a VLM for item classification, running both locally on edge. Key elements: gvadetect, gvagenai Models: yolo26s, MiniCPM-V-4_5
	ONVIF Camera Analytics Validation Python Use a VLM as an additional validation layer for ONVIF-enabled analytics cameras. Key elements: gvagenai Models: Configurable VLM
	Image Embeddings Generation with ViT CLI Generate image embeddings using the Vision Transformer component of a CLIP model. Key elements: gvainference Models: clip-vit-large-patch14 (CLIP ViT)

Audio analytics

Preview	Sample
	Audio Event Detection CLI Audio event detection, converting results to JSON. Key elements: gvaaudiodetect, gvametaconvert, gvametapublish Models: aclnet
	Audio Transcription CLI Speech transcription using an OpenVINO GenAI Whisper model. Key elements: gvaaudiotranscribe Models: whisper

3D: LiDAR & radar

Preview	Sample
	PointPillars Inference with g3dinference CLI Complete LiDAR-only 3D detection pipeline. Key elements: g3dlidarparse, g3dinference Models: PointPillars
	LiDAR Parse CLI LiDAR parsing pipeline. Key elements: g3dlidarparse Models: — (parsing only)
	Live LiDAR Capture CLI Real-time LiDAR capture from a physical device (RoboSense via rs_driver). Key elements: g3dlidarsrc, g3dinference Models: PointPillars
	Camera + 3D Object Fusion CLI Fuse 2D camera detections with 3D LiDAR detections. Key elements: g3dobjectfuser, gvastreammux Models: yolo11n, PointPillars
	Radar Signal Process CLI mmWave radar signal processing with point-cloud detection, clustering and tracking. Key elements: g3dradarprocess Models: — (signal processing)

Cameras & input sources

Preview	Sample
	RealSense™ Camera CLI Capture a video stream from a 3D Intel RealSense™ Depth Camera. Key elements: gvarealsense Models: — (capture only)
	ONVIF Camera Discovery Python Automatically discover ONVIF cameras on the network and launch pipelines for each. Key elements: gvadetect Models: Configurable detector
	Multi-camera deployments CLI Handle video streams from multiple cameras in a single application. Key elements: gvadetect, gvafpscounter Models: yolo11s (many YOLO variants)
	Multi-Stream Mux/Demux CLI Share a single inference pipeline across streams with per-source routing. Key elements: gvastreammux, gvastreamdemux Models: Configurable detector

Metadata: publishing, access & visualization

Preview	Sample
	Metadata Publishing CLI Convert inference metadata to JSON and publish to file or Kafka/MQTT. Key elements: gvametaconvert, gvametapublish Models: centerface, dima806_fairface_gender_image_detection, dima806_facial_age_image_detection
	gvaattachroi CLI Define the regions on which inference should be performed. Key elements: gvaattachroi, gvadetect Models: yolov8s
	FPS Throttle CLI Throttle framerate independently of sink sync, without frame duplication or dropping. Key elements: gvafpsthrottle Models: —
	Watermark Metadata Python Attach custom drawing primitives (hexagons, lines, circles, text) and render them. Key elements: gvawatermark Models: — (drawing only)
	Draw Face Attributes (C++) C++ Set a C callback to access frame metadata and visualize inference results. Key elements: gvadetect, gvaclassify Models: centerface, dima806_facial_age_image_detection, dima806_fairface_gender_image_detection, dima806_face_emotions_image_detection
	Draw Face Attributes (Python) Python Set a Python callback to access frame metadata and visualize inference results. Key elements: gvadetect, gvaclassify Models: centerface, dima806_facial_age_image_detection, dima806_fairface_gender_image_detection, dima806_face_emotions_image_detection
	Open Close Valve Python Open/close a GStreamer valve branch from a callback based on detection results. Key elements: gvadetect, valve Models: yolo11s, dima806_vehicle_10_types_image_detection
	Hello DL Streamer Python Build a detection pipeline, analyze metadata to count objects, and visualize results. Key elements: gvadetect, gvawatermark Models: yolo11n

Customization & extensibility

Preview	Sample
	Custom Post-Processing Library — Classification CLI, C++ Write a custom post-processing library that converts emotion-classification outputs to GstAnalytics metadata. Key elements: gvaclassify Models: centerface, hsemotion
	Custom Post-Processing Library — Detection CLI, C++ Write a custom post-processing library that converts YOLOv11 tensor outputs to detection metadata. Key elements: gvadetect Models: yolo11s
	gvapython — Face Detection and Classification CLI, Python Customize a pipeline with a Python script for inference post-processing. Key elements: gvapython, gvadetect, gvaclassify Models: centerface, dima806_fairface_gender_image_detection, dima806_facial_age_image_detection
	gvapython — Save Frames with ROI CLI, Python Use gvapython to save video frames containing detected objects to disk. Key elements: gvapython, gvadetect Models: centerface
	python-elements — Face Detection and Classification CLI, Python Build a custom Python GStreamer element using the GstAnalytics metadata API. Key elements: gvaagelogger_py, gvadetect, gvaclassify Models: YOLOv8-Face-Detection, fairface_age_image_detection, fairface_gender_image_detection
	python-elements — Save Frames with ROI CLI, Python Build a custom Python GStreamer element to save frames with detected objects. Key elements: gvaframesaver_py, gvadetect Models: YOLOv8-Face-Detection
	python-elements — Loitering Detection CLI, Python Measure object dwell time with a custom Python element and render a visual alert when the threshold is exceeded. Key elements: gvaanalytics, gvawatermark Models: yolo11s
	Face Detection and Classification (Python) Python Download models from Hugging Face, export to OpenVINO IR, and run inference. Key elements: gvadetect, gvaclassify Models: YOLOv8-Face-Detection, fairface

Performance & benchmarking

Preview	Sample
	Benchmark CLI, Python Measure the performance of single- or multi-channel video analytics pipelines. Key elements: gvadetect, gvafpscounter Models: centerface (configurable)
	DL Streamer E2E Performance Python Compare DL Streamer vs. OpenCV + OpenVINO throughput with a YOLO26s INT8 model. Key elements: gvadetect Models: yolo26s (INT8)

Interoperability

Preview	Sample
	DL Streamer and DeepStream Coexistence Python Run pipelines on DL Streamer and/or NVIDIA DeepStream side by side. Key elements: gvadetect Models: yolov8 license-plate detector, PP-OCRv4
	Coexistence Benchmark Python Measure the maximum number of concurrent LPR streams on systems combining Intel and NVIDIA hardware. Key elements: gvadetect Models: yolov8 license-plate detector, PP-OCRv4

Auto-generated reference applications

These end-to-end reference apps combine multiple elements into complete solutions. Find them under samples/auto_generated_samples.

Preview	Sample
	DeepStream Test4 → DL Streamer Conversion CLI DL Streamer equivalent of NVIDIA’s deepstream-test4 with YOLO11n detection and metadata publishing. Key elements: gvadetect, gvametaconvert, gvametapublish Models: yolo11n
	DeepStream LPR App Conversion (C++) C++ C++ conversion of NVIDIA’s DeepStream LPR app — license plate detection, tracking and text recognition. Key elements: gvadetect, gvatrack, gvaclassify Models: YOLOv11, PaddleOCR
	License Plate Recognition CLI Detect license plates with YOLOv11 and recognize text with PaddleOCR. Key elements: gvadetect, gvainference Models: YOLOv11, PaddleOCR
	Multi-Stream Compose CLI Multi-camera analytics with composite WebRTC output, on-demand recording and a 2x2 GPU-accelerated mosaic. Key elements: gvadetect, gvastreammux, gvawatermark Models: yolo11s
	People Detection and Tracking with Deep SORT CLI Detect and track people using YOLO26m and Deep SORT with a Mars-Small-128 re-ID model. Key elements: gvadetect, gvatrack Models: yolo26m, mars-small128
	Pose Estimation Compose CLI Run 4 YOLO pose models in parallel on the same video and composite results into a 2x2 mosaic. Key elements: gvaclassify, gvawatermark Models: yolo26n-pose, yolo11n-pose, yolov8n-pose, yolov8l-pose
	Safety Compliance Monitor CLI Detect and track workers and use Qwen2.5-VL to verify helmet and harness compliance. Key elements: gvadetect, gvatrack, gvagenai Models: yolo26m, Qwen2.5-VL-3B
	Smart NVR — Event-Based Recording CLI Detect people with YOLO11n and record video only when a person is present. Key elements: gvadetect Models: yolo11n