Skip to main content
Ctrl+K

Open Edge Platform

    • Open Edge Platform
    • Metro
    • Manufacturing
    • Retail
    • Robotics
    • Education
    • Health and Life Sciences
    • Federal and Aerospace
    • Libraries, Tools, Services
    • Edge Microvisor Toolkit
    • Image Composer Tool
  • Open Edge Platform
  • Metro
  • Manufacturing
  • Retail
  • Robotics
  • Education
  • Health and Life Sciences
  • Federal and Aerospace
  • Libraries, Tools, Services
  • Edge Microvisor Toolkit
  • Image Composer Tool

Section Navigation

Tools

  • VIPPET
    • Get Started
      • Installation
        • System Requirements
        • Pre-Installation Steps
        • Use Pre-Built Docker Images
        • Build from Source
        • Installation Troubleshooting
      • Quickstart Guide
        • Vision Use Case
        • GenAI Use Case
        • Motion Detection Use Case
      • Get Support
    • User Guide
      • Input Management
        • Cameras
        • Image Sets
        • Videos
      • Model Management
        • Using Models from Geti
        • Using Models from Hugging Face
      • Working With Pipelines
        • Creating New Pipelines
        • Configuring and Running Pipelines
        • Using Predefined Pipelines
        • Simple vs. Advanced View
      • Benchmarking
        • Performance Testing
        • Stream Density Testing
        • Managing Jobs
    • Developer Guide
      • Architecture
        • VIPPET UI
        • VIPPET Backend
      • Performance Metrics
        • Pipeline Performance
        • System Performance
      • Contributing Guide
        • Backend
        • How to add a new pipeline
        • How to add a new element
    • Troubleshooting
    • Release Notes
      • Release Notes 2026.1
      • Release Notes 2026.0
      • Release Notes 2025.2
  • Scenescape
    • Get Started
      • System Requirements
      • Prerequisites
    • How to Guides
      • Deploy Scenescape
      • Use the UI and Online Documentation
      • Build a Scene
        • Create a New Scene
        • Generate a Scene Map
        • Use Different Sensor Types
        • Visualize ROIs and Regions
        • Configure a Hierarchy of Scenes
        • Configure Geospatial Coordinates
        • Configure Geospatial Map Service API Keys
        • Configure Spatial Analytics
      • Integrate Cameras and Sensors
      • Calibrate Cameras
        • Use 2D UI for Calibration
        • Use 3D UI for Calibration
        • Automatic Calibration - Visual Features
        • Automatic Calibration - April Tags
      • Work with Spatial Analytics Data
    • Other Topics
      • Defining Object Properties
      • Enabling Re-identification
      • Integrating Intel® Geti™ AI Models
      • Configuring DL Streamer Video Pipeline
      • Model configuration file format
      • Running License Plate Recognition with 3D Object Detection
      • Managing Files in Volumes
      • Controlling Scene Lighting with Physical Light Sensors
    • Additional Resources
      • Scenescape Hardening Guide
      • How to Upgrade Scenescape
      • How Scenescape converts Pixel-Based Bounding Boxes to Normalized Image Space
    • Microservices
      • Auto Camera Calibration
        • Get Started
          • Build from Source
        • API Reference
        • Markerless Camera Calibration Internals
      • Cluster Analytics
        • Get Started
          • How to Build Cluster Analytics from Source
      • Scene Controller
        • Get Started with Scene Controller
          • How to Build Scene Controller from Source
        • How to Configure the Tracker
        • 2-Tier Hybrid Search Implementation
        • API Reference
        • Scene Controller Data Formats
        • Pose Adjustment Package Design
      • Mapping Service
        • How to Build from Source
    • API Reference
    • Troubleshooting
    • Release Notes
      • Release Notes 2025
  • Geti
  • Geti Instant Learn

Libraries

  • Anomalib
  • Datumaro
  • DL Streamer
    • Get Started
    • System Requirements
    • Install Guide
      • Install Guide Ubuntu
      • Install Guide Ubuntu 24.04 on WSL2
      • Uninstall Guide Ubuntu
      • Install Guide Windows
      • Uninstall Guide Windows
    • Tutorial
    • Samples
    • Supported Models
    • Elements
      • gvadetect
      • gvaclassify
      • gvainference
      • gvatrack
      • gvaaudiodetect
      • gvaaudiotranscribe
      • gvagenai
      • g3dradarprocess
      • g3dlidarsrc
      • g3dlidarparse
      • g3dinference
      • g3dobjectfuser
      • g3drender
      • gvaanalytics
      • gvaattachroi
      • gvafpscounter
      • gvafpsthrottle
      • gvastreammux
      • gvastreamdemux
      • gvametaaggregate
      • gvametaconvert
      • gvametapublish
      • gvapython
      • gvarealsense
      • gvawatermark
      • gvamotiondetect
      • GStreamer Elements
        • Compositor
    • Developer Guide
      • Coding Agent
      • Advanced Installation Guide
        • Advanced Installation On Ubuntu - Prerequisites
        • Advanced Installation - Compilation From Source
        • Advanced Installation On Ubuntu - Build Docker Image
        • Advanced Uninstallation On Ubuntu
        • Advanced Installation on Windows - Compilation From Source
        • Advanced Installation on Windows - Install via Command Line
      • Metadata
        • GStreamer Analytics Metadata
        • 3D Sensor Metadata (LiDAR & radar)
        • Watermark Metadata
        • Legacy Analytics Metadata (deprecated)
      • Model Preparation
        • YOLO Models
        • Transformer Models
        • Download Public Models
      • OpenVINO Custom Operations Support
      • Model Info Section
      • GStreamer Python Bindings
      • Custom GStreamer Plugin Installation
      • Custom Processing
      • Object Tracking
      • GPU device selection
      • Heterogeneous AI Inference on Intel Core Ultra processors
      • Performance Guide
      • Profiling with Intel VTune™
      • DL Streamer and DeepStream Coexistence
      • Converting NVIDIA DeepStream Pipelines to Deep Learning Streamer Pipeline Framework
      • How to Contribute
        • Coding Style
      • Latency Tracer
      • Model-proc File (legacy)
        • How to Create Model-proc File
      • Optimizer
    • API Reference
    • Architecture 2.0
      • Migration to 2.0
      • Memory Interop and C++ abstract interfaces
      • ② C++ elements
      • ③ GStreamer Elements
      • ③ GStreamer Bin Elements
      • Python Bindings
      • PyTorch tensor inference
      • Elements 2.0
      • Packaging
      • Samples 2.0
      • API 2.0 Reference
    • Release Notes: Deep Learning Streamer (DL Streamer) Pipeline Framework Release 2026.1
  • PLCopen Motion Control
    • RTmotion Library
      • Installation & Setup
        • System Requirements
        • OS Setup
        • Real-Time in Linux
      • RTmotion Concept and Application Interface
    • Notices and Disclaimers
  • Geti SDK
  • EtherCAT Master Stack
  • Robot Motion Control Task
  • Video Chunking Utils
    • Release Notes

Microservices

  • AAgent Quality Handler
    • Get Started
      • System Requirements
    • How It Works
    • Build From Source
    • How to Integrate
    • API Reference
    • Troubleshooting
    • Release Notes
  • Audio Analyzer
    • Get Started
      • System Requirements
      • Configuration
      • Build From Source
      • Run With Docker Compose
      • Run On the Host
    • How It Works
    • API Reference
    • Troubleshooting
    • Release Notes
  • DL Streamer Pipeline Server
    • Get Started
      • System Requirements
      • Environment Variables
      • Build from Source
      • Deploy with Helm
    • How-to Guides
      • Manage Pipeline
      • Autostart Pipelines
      • Change Deep Learning Streamer Pipeline
      • Run Configurable Pipelines
      • Run UDF Pipelines
      • Use CPU for Inference
      • Use GPU or NPU for Inference
      • Use Image File as Source over REST Payload
      • Use RTSP Camera as Source
      • Download and Run YOLO Models
      • Publish Frames to S3 Storage
      • Publish Data over MQTT
      • Publish Metadata to InfluxDB
      • Stream Frames over WebRTC
      • Publish Metadata over ROS2
      • Add System Timestamps to Metadata
    • Advanced User Guide
      • Basic Deep Learning Streamer Pipeline Server Configuration
      • REST API guide
        • REST Endpoints Reference Guide
        • Defining Media Analytics Pipelines
        • Customizing Pipeline Requests
      • Cameras
        • RTSP Cameras
      • File Ingestion
        • Image Ingestion
        • Video Ingestion
        • Multifilesrc Usage
      • User Defined Functions (UDF)
        • UDF Writing Guide
        • Configuring udfloader element
      • Publishers
        • MQTT Publishing via gvapython
        • MQTT Publishing
        • OPCUA Publishing post pipeline execution
        • S3 Frame Storage
      • How To Advanced
        • Object tracking with UDF
        • Enable HTTPS for DL Streamer Pipeline Server (Optional)
        • Performance Analysis (Latency)
        • Pinning the DL Streamer Pipeline Sever to CPU cores
        • Get tensor vector data
        • Run multistream pipelines with shared model instance
        • Cross stream batching
        • Enable Open Telemetry
        • Working with other services
    • API Reference
    • Troubleshooting
    • Release Notes
      • Release Notes 2025
      • Release Notes 2024
  • Document Ingestion - PGVector
    • Get Started
      • System Requirements
    • Build and customize options
    • API Reference
    • Release Notes
  • Metrics Manager
    • Get Started
      • System Requirements
      • Building from Source
      • Helm Deployment
      • Environment Variables
      • Custom Metrics Scripts
      • Testing Guide
    • How It Works
    • API Reference
    • Troubleshooting
    • Release Notes
  • Model Download
    • Get Started
      • Migrate from Model Registry
      • System Requirements
      • Ephemeral Container
      • Build from Source
      • Deploy with Helm Chart
    • Run Unit Tests
    • API Reference
    • Release Notes
      • Release Notes 2025
  • Multimodal Embedding Serving
    • Get Started
      • System Requirements
      • Build from Source
    • SDK Usage Guide
    • Wheel-Based Installation Guide
    • Supported Models
    • API Reference
    • Release Notes
      • Release Notes 2025
  • Text To Speech
    • Get Started
      • System Requirements
      • Configuration
      • Run With Docker Compose
    • How-to Guides
      • Run On the Host
      • Build From Source
    • How It Works
    • API Reference
    • Troubleshooting
    • Release Notes
  • Time Series Analytics
    • Get Started
      • System Requirements
      • Deploy with Helm
    • How It Works
    • Access Microservice API
    • Configure Microservice
    • API Reference
    • Release Notes
      • Release Notes 2025
  • Vector Retriever - milvus
    • Get Started Guide
      • System Requirements
    • API Reference
    • Release Notes
      • Release Notes 2025
  • Visual Data Preparation For Retrieval
  • Multi-level Video Understanding
    • Get Started
      • System Requirements
      • Build from Source
      • Adding Swap Space
    • API Reference
    • Release Notes
      • Release Notes 2025

Sample Applications

  • Chat Q&A
    • Get Started
      • System Requirements
      • Build from Source
      • Deploy with Helm
    • How It Works
    • How to Test Performance
    • Benchmarks
    • API Reference
    • Release Notes
      • Release Notes 2025
  • Chat Q&A Core
    • Get Started
      • System Requirements
    • How to Build from Source
    • How to deploy with Helm
    • Benchmarks
    • API Reference
    • Release Notes
      • Release Notes 2025
  • Document Summarization
    • Get Started
      • System Requirements
    • Architecture
    • How to Build from Source
    • How to deploy with Helm
    • How to Test Performance
    • API Reference
    • Troubleshooting
    • Release Notes
      • Release Notes 2025
  • Video Search and Summarization
    • Get Started
      • System Requirements
    • How It Works
      • Video Search
      • Video Summarization
      • Video Search and Summarization
    • How to Build from Source
    • How to deploy with Helm* Chart
    • Deploy VSS with vLLM
    • Directory Watcher Service Guide
    • API Reference
    • MCP Server for VSS
    • Troubleshooting
    • Release Notes
      • Release Notes 2025

Model Deployment

  • OpenVINO
  • OpenVINO Model Server

---------------

  • Intel® Edge System Qualification
  • Get Help or Contribute
  • Edge AI Libraries
  • Scenescape
  • Other Topics
  • How to Enable Re-identification Using Visual Similarity Search

How to Enable Re-identification Using Visual Similarity Search#

This guide provides step-by-step instructions to enable or disable re-identification (ReID) using visual similarity search in a Scenescape deployment. By completing this guide, you will:

  • Enable re-identification using a visual database and feature-matching model.

  • Understand how to track and evaluate unique object identities across frames.

  • Learn how to tune performance for specific use cases.

This task is important for enabling persistent object tracking across different camera scenes or time intervals.


Prerequisites for Re-identification#

Before you begin, ensure the following:

  • Docker is installed and configured.

  • You have access to modify the docker-compose.yml file in your deployment.

  • You are familiar with scene and camera configuration in Scenescape.


Steps to Enable Reidentification (ReID) for Out of Box Experience#

  1. Select one ReID database

    Use one backend override with the base Compose file. Both overrides create the same logical reid service and configure the Scene Controller. Run these commands from the sample_data/ directory:

    # VDMS
    docker compose -f docker-compose-dl-streamer-example.yml \
      -f docker-compose.vdms-override.yml \
      --profile controller up
    
    # Or Qdrant
    docker compose -f docker-compose-dl-streamer-example.yml \
      -f docker-compose.qdrant-override.yml \
      --profile controller up
    

    Use exactly one override. Do not combine them. No backend-specific Compose profile or manual service/dependency editing is required.

    From the repository root, make demo-reid starts the core demo plus the ReID database, defaulting to VDMS. Switch backends with REID_BACKEND:

    make demo-reid
    make demo-reid REID_BACKEND=qdrant
    

    Plain make demo runs tracking without ReID. make demo-close uses the backend override recorded when the demo was started.

  2. Enable Visual Feature Extraction in Video Pipeline Edit the retail-config setting in Docker Compose as follows:

retail-config:
  file: ./dlstreamer-pipeline-server/retail-config-reid.json

This reidentification-specific configuration uses a vision pipeline that includes anonymous visual feature extraction (also called “visual embeddings”) using a person reidentification model:

"pipeline": "multifilesrc loop=TRUE location=/home/pipeline-server/videos/apriltag-cam2.ts name=source ! decodebin ! videoconvert ! video/x-raw,format=BGR ! sscape_timestamp_capture name=timesync ntp-server=ntpserv use-frame-ntp-timestamp=false ! gvadetect model=/home/pipeline-server/models/intel/person-detection-retail-0013/FP32/person-detection-retail-0013.xml model-proc=/home/pipeline-server/models/object_detection/person/person-detection-retail-0013.json name=detection ! gvainference model=/home/pipeline-server/models/intel/person-reidentification-retail-0277/FP32/person-reidentification-retail-0277.xml inference-region=roi-list ! gvametaconvert add-tensor-data=true name=metaconvert ! sscape_post_inference_data_publish name=datapublisher ! gvametapublish name=destination ! appsink sync=true",

Expected Result: Scenescape starts with ReID enabled and begins assigning UUIDs based on visual similarity.


Selecting the ReID Vector Database Backend#

VDMS and Qdrant are mutually exclusive alternatives. They use the same service name (reid), hostname (reid.scenescape.intel.com), port (55555), TLS material, and controller connection settings. The selected override sets REID_DATABASE and starts the matching database implementation.

Prerequisites#

  • ReID is already enabled (feature extraction pipeline and reid-config.json as in the steps above).

  • Secrets include shared ReID certificates (scenescape-reid* / scenescape-reid-s*). Regenerate with make clean-secrets && make init-secrets if those files are missing.

  • You can pass an override file when starting services.

Steps#

  1. Stop the stack using the same base and backend override files used to start it:

    docker compose -f docker-compose-dl-streamer-example.yml \
      -f docker-compose.vdms-override.yml \
      --profile controller down
    
  2. Start with the other backend override. For example, to select Qdrant:

    docker compose -f docker-compose-dl-streamer-example.yml \
      -f docker-compose.qdrant-override.yml \
      --profile controller up
    

    The override (docker-compose.qdrant-override.yml):

    • Starts the logical reid service using Qdrant, with TLS on shared host reid.scenescape.intel.com and port 55555

    • Sets REID_DATABASE=QDRANT on the scene service

    • Connection defaults (hostname, port, TLS=true, cert paths) are shared via REID_* settings

  3. Do not combine the backend override files. A deployment has one logical reid service and one selected adapter.

Shared ReID environment variables#

Only REID_DATABASE selects the backend. Connection and tuning use shared REID_* names (adapters ignore knobs they do not need). Hostname, port, TLS, and certificate paths are the same for every backend.

Variable

Purpose

Default

REID_DATABASE

Backend selector (VDMS or QDRANT)

VDMS

REID_HOSTNAME

Database host

reid.scenescape.intel.com

REID_PORT

Database port (1–65535)

55555

REID_USE_TLS

Use TLS (true/false)

true

REID_API_KEY

Optional API key

unset (Qdrant)

REID_CONFIDENCE_THRESHOLD

TIER 1 metadata confidence threshold

0.8

REID_CA_CERT / REID_CLIENT_CERT / REID_CLIENT_KEY

TLS / mTLS material

/run/secrets/certs/scenescape-ca.pem, scenescape-reid.crt, scenescape-reid.key

Backend-prefixed names such as VDMS_HOSTNAME or QDRANT_PORT are no longer read. Set the REID_* equivalent instead.

Values are validated at controller startup. A port outside 1–65535, a confidence threshold outside 0.0–1.0, or an unrecognized boolean stops the controller with a message naming the variable and its value, so a typo cannot silently disable TLS or widen a threshold.

Switching back to VDMS#

  1. Stop the Qdrant-backed stack.

  2. Replace the Qdrant override with the VDMS override:

    docker compose -f docker-compose-dl-streamer-example.yml \
      -f docker-compose.vdms-override.yml \
      --profile controller up
    

Note: Vector data is not migrated between VDMS and Qdrant. After a backend switch, identities are matched only against embeddings stored in the newly selected database.

Kubernetes (Helm)#

The Helm chart mirrors the Compose model: a single logical reid Service backed by exactly one database Deployment, sharing the reid.scenescape.intel.com certificates and port 55555. Select the backend with reid.backend:

helm upgrade scenescape-release-1 --install kubernetes/scenescape-chart/ \
  -n scenescape --create-namespace \
  --set reid.enabled=true --set reid.backend=qdrant

The chart sets REID_DATABASE on the Scene Controller from reid.backend, so no other value needs to change. Setting reid.enabled=false removes the database Deployment, Service, and ReID certificates, and drops the ReID client certificates from the Scene Controller.

From the repository root, make demo-k8s follows the same tiers as the Compose demo:

make demo-k8s                                        # core services, no ReID
make demo-k8s DEMO_K8S_MODE=reid                     # core plus ReID (VDMS)
make demo-k8s DEMO_K8S_MODE=reid REID_BACKEND=qdrant # core plus ReID (Qdrant)
make demo-k8s DEMO_K8S_MODE=all                      # ReID plus mapping and cluster analytics

Expected Result: The Scene Controller connects to Qdrant, creates or verifies the ReID collection, and continues UUID assignment via visual similarity.

Service-link environment variables#

The ReID Service is named reid, so Kubernetes injects REID_PORT=tcp://<clusterIP>:<port> into every pod in the namespace, which collides with the REID_PORT setting described above. The chart sets enableServiceLinks: false on the Scene Controller to suppress this; all of its dependencies are addressed by DNS.

If you write your own manifests, either do the same or set REID_PORT explicitly, since values in env take precedence over service links. As a backstop, the controller ignores any REID_* value that looks like a service link (tcp://…) and logs a warning rather than failing to start.

ReID pod filesystem#

The ReID container runs with readOnlyRootFilesystem: true. Each backend gets emptyDir volumes for the only paths it writes:

Backend

Writable mounts

Notes

VDMS

/vdms/data, /tmp

OVERRIDE_db_root_path moves the database off the image layer

Qdrant

/qdrant/storage, /qdrant/snapshots

QDRANT_INIT_FILE_PATH moves the init indicator into writable storage

Because the VDMS image writes its generated config next to the server binary, the chart renders that config into /vdms/data and starts the server with -cfg. If you pin a different VDMS image, confirm it still provides override_default_config.py and the -cfg flag.

ReID vector data is stored in emptyDir and is lost when the pod restarts, which matches the behaviour before the volumes existed. Replace reid-data with a PersistentVolumeClaim if the embeddings must survive restarts.

The pod still runs as root (runAsUser: 0) because both upstream images expect it; that is a separate hardening step.


Steps to Disable Re-identification#

  1. Stop using the backend override

    Stop the stack with its active override, then restart the base Compose file without either ReID backend override. The base file does not contain a ReID database service.

    docker compose -f docker-compose-dl-streamer-example.yml \
      -f docker-compose.vdms-override.yml \
      --profile controller down
    

    Substitute docker-compose.qdrant-override.yml when Qdrant is active.

  2. Remove ReID from the Camera Pipeline Edit the retail-config setting in Docker Compose and revert to the config without re-id model:

retail-config:
  file: ./dlstreamer-pipeline-server/retail-config.json
  1. Restart the System:

    docker compose --profile controller up --build
    

Expected Result: Scenescape runs without ReID and no visual feature matching is performed.


Evaluating Re-identification Performance#

  • Track Unique IDs:
    Scenescape publishes unique_detection_count via MQTT under the scene category topic. Each object includes an id field (UUID) for tracking.

  • UI Support:
    UUID display in the 3D UI is planned for future releases.

Note: The default ReID model is tuned for the ‘person’ category and may not generalize well to other object types.


How Re-identification Works#

When an object is first detected, it is assigned a UUID and no similarity score. If ReID is enabled, the system collects visual features over time. Once enough features are gathered, they are compared to those in the database:

  • Match Found: The object is reassigned a matching UUID and given a similarity score.

  • No Match: The object retains its original UUID.

The scene output includes reid_state for each tracked object. For canonical state definitions and lifecycle transitions, see 2-Tier Hybrid Search Implementation. For output field contract details, see Scene Controller Data Formats.

Known Issue: Current VDMS implementation does not support feature expiration, leading to degraded performance over time. This will be addressed in a future release.


Configuration Options#

Parameter

Purpose

Expected Value/Range

DEFAULT_SIMILARITY_THRESHOLD_L2 / DEFAULT_SIMILARITY_THRESHOLD_COSINE

Match-acceptance threshold defaults selected by similarity_metric: the default COSINE metric uses 0.5; explicitly configured L2 uses 40.0.

Float; tune per metric. For COSINE/IP, values such as 0.2–0.8 may be used. For L2, use a distance threshold appropriate to the embedding/model.

DEFAULT_MINIMUM_BBOX_AREA

Minimum bounding box size to consider a valid feature.

Pixel area (e.g., 400–1600)

DEFAULT_MINIMUM_FEATURE_COUNT

Minimum features needed before querying DB.

Integer (e.g., 5–20)

DEFAULT_MAX_FEATURE_SLICE_SIZE

Proportion of features stored to improve DB performance.

Float (e.g., 0.1–1.0)

To apply changes, use the same backend override you selected when starting the stack:

docker compose -f docker-compose-dl-streamer-example.yml \
  -f docker-compose.vdms-override.yml \
  --profile controller down
make -C docker
docker compose -f docker-compose-dl-streamer-example.yml \
  -f docker-compose.vdms-override.yml \
  --profile controller up --build

Troubleshooting#

  1. Issue: ReID not working

    • Cause: Database container is not running, not linked, or TLS/certs do not match the shared ReID defaults.

    • Resolution:

      docker compose -f docker-compose-dl-streamer-example.yml \
        -f docker-compose.vdms-override.yml \
        --profile controller ps reid
      docker compose -f docker-compose-dl-streamer-example.yml \
        -f docker-compose.vdms-override.yml \
        --profile controller logs reid
      

      Substitute the Qdrant override when it is selected. Confirm the reid service is healthy, the expected REID_DATABASE is set on scene, and the shared scenescape-reid* certificates exist.

  2. Issue: Objects not re-identifying across scenes

    • Cause: Insufficient visual features collected or poor lighting.

    • Resolution:

      • Lower DEFAULT_MINIMUM_FEATURE_COUNT.

      • Increase DEFAULT_MINIMUM_BBOX_AREA only if objects are large and visible.

  3. Issue: Backend switch appears to “lose” identities

    • Cause: VDMS and Qdrant do not share stored embeddings.

    • Resolution: Expected after switching REID_DATABASE. Re-accumulate features in the new backend, or restore the previous backend and its data volume.

On this page
  • Prerequisites for Re-identification
  • Steps to Enable Reidentification (ReID) for Out of Box Experience
  • Selecting the ReID Vector Database Backend
    • Prerequisites
    • Steps
    • Shared ReID environment variables
    • Switching back to VDMS
    • Kubernetes (Helm)
      • Service-link environment variables
      • ReID pod filesystem
  • Steps to Disable Re-identification
  • Evaluating Re-identification Performance
  • How Re-identification Works
  • Configuration Options
  • Troubleshooting

This Page

  • Show Source
Enable cookies to use AI chat
Chat is locked. Click the blue chat button, then enable Functional cookies to unlock it.