OMS Reference Document
  • Welcome to OMS
  • Estimation
    • Why estimate?
    • Login and set-up
    • Accessing cameras
    • Performing the task
    • Interactive and time management
    • Calibration documents
      • Aquatica Orlando (APO)
      • SeaWorld Orlando (SWO)
      • Busch Gardens Tampa Bay (BGT)
      • Sesame Place (SP)
      • Busch Gardens Williamsburg (BGW)
      • Aquatica San Antonio, Texas (APT)
      • SeaWorld San Antonio, Texas (SWT)
      • SeaWorld San Diego, California (SWC)
    • Supervisor
  • Verification
    • Why verification?
    • Login and set-up
    • Performing the task
      • People Search Method
    • Supervisor
  • Scan Audit
    • Why scan audit?
    • Login and set-up
      • Internal document
      • Navigating Asana
    • Smart and Final
    • Heinen's checkout audit
    • Supervisor
  • Screener+
    • Why monitor alerts?
    • Login and set-up
    • Performing the task
      • Morning star
      • Nirvana
      • Tractor Supply Co
      • GAP Inc
        • Disable Code
    • Supervisor
  • Configuration Audit
    • Login and Set-up
    • Simon entrance audit
    • License and Storage Configuration
    • Ankgor wine and liquor camera mapping
    • Sensormatic dwell time
    • Supervisor
  • CVAT
    • Why CVAT?
    • Login and set up
    • Performing the task
      • Tasks
    • Supervisor
  • Infrastructure Monitoring
    • Login and set-up
    • Performing the task
    • Supervisor
  • Installations
    • Performing the task
    • Supervisor
  • Supervisor responsibilities
Powered by GitBook
On this page
  • Hand and Object
  • Smoke and Fire
  • People and Vehicle
  • Safety Vest and Hat

Was this helpful?

  1. CVAT
  2. Performing the task

Tasks

PreviousPerforming the taskNextSupervisor

Last updated 5 months ago

Was this helpful?

In CVAT, various annotation tasks are performed to help train and refine AI algorithms for specific scenarios. Here is an outline of the four main annotation tasks currently in use:

Hand and Object

  • Description: This task involves annotating hands and the objects they interact with during scanning at a supermarket checkout.

  • Objective: To precisely identify the interaction between hands and items being scanned to train AI models for object-hand interaction.

  • Steps:

    • Annotate hands visible in the frame.

    • Mark objects being scanned or handled by the hands.

    • Ensure bounding boxes accurately cover the hand and the associated object.

Smoke and Fire

  • Description: This task focuses on identifying and annotating smoke or fire in security camera footage.

  • Objective: To help detect potential fire hazards by teaching AI models to recognize smoke and fire patterns in real-time.

  • Steps:

    • Identify frames where smoke or fire is visible.

    • Draw bounding boxes around areas with visible smoke or flames.

    • Ensure annotations include subtle smoke to improve detection accuracy.

People and Vehicle

  • Description: This task involves annotating people and vehicles captured in camera footage, typically in car parks or similar environments.

  • Objective: To differentiate between people and vehicles and improve AI capabilities in monitoring and surveillance tasks.

  • Steps:

    • Annotate each person visible in the frame with a bounding box.

    • Annotate each vehicle (e.g., cars, bikes, trucks) in the frame.

    • Distinguish between categories (people vs. vehicles) to ensure clarity.

Safety Vest and Hat

  • Description: This task involves annotating safety vests and helmets worn by individuals in construction sites or loading docks.

  • Objective: To ensure compliance with safety regulations by detecting individuals without proper safety gear.

  • Steps:

    • Annotate individuals wearing safety vests.

    • Annotate individuals wearing safety hats.

    • Note individuals who lack either a vest or a helmet.


Conclusion

Each task plays a crucial role in developing robust AI systems, whether for enhancing retail efficiency, improving safety compliance, or enabling advanced surveillance. By adhering to these guidelines, we ensure high-quality annotations that contribute significantly to AI training and development.