Tasks
Last updated
Was this helpful?
Last updated
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:
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.
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.
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.
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.
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.