Review movement during real tasks
A camera that looks stable while a person stands still can behave differently when they bend, walk, or reach overhead. Inspect complete task examples that include those movements. Note whether the active work area remains interpretable and whether rapid direction changes hide transitions. A visually dramatic recording is not necessarily unusable; suitability depends on the task being modeled.
Distinguish sources of image change
Head turns, mount vibration, motion blur, and stabilization artifacts create different problems. Identify which one is responsible before changing the specification. A tighter mount may address vibration without changing intentional viewpoint motion. Processing that reduces motion can also crop the image, so inspect whether hands or objects disappear at the newly created frame boundaries.
Document processing history
Ask whether delivered footage is raw, stabilized, cropped, or otherwise transformed. Preserve a source identifier and describe any version changes. Labels and camera information need to correspond to the actual delivered pixels. If both original and processed video are requested, define how they relate rather than assuming their frame counts and timestamps remain interchangeable.
Set task-specific acceptance rules
For activity classification, a brief blurred transition may leave adequate context. For small object localization, the same interval may remove required evidence. Write examples of acceptable and unacceptable motion for the intended task, then review a varied pilot against them. Avoid promising that a single camera setting will solve every visibility problem across environments and participants.
Review checklist
- Review ordinary body movement
- Identify blur, vibration, and cropping
- Record stabilization and transformations
- Use task-specific acceptance examples