Skip to main content

Stepwise Mode

Stepwise mode is designed for more precise decisions using both detection steps and exception conditions.

Best Use Cases​

  • When simple object presence causes frequent false positives
  • When you need rule-based logic like "this must be true, and this must not be true"


Example Inputs​

  • Detect people using mobile phones
  • Alert when workers stay in a restricted area for too long


Template Tips​

  • Save recurring rules (for example, phone use or restricted-area stay) as templates.
  • For scenarios with many exceptions, keep a master template and clone it per camera.
  • Validate template updates on representative cameras before broad rollout.


Writing Tips​

Detailed logic usually has two blocks.

  • steps: conditions that should trigger alerts
  • exceptions: conditions that should suppress alerts

Operational tips:

  • Start with 2-3 core conditions in steps
  • Add exceptions based on frequently observed false-positive cases
  • Change one item at a time and verify results


How To Improve Decision Quality​

Stepwise mode works best when detection and exception logic are clearly separated.

  • Write steps as broad suspicious-condition criteria.
  • Write exceptions as clear normal-condition filters.
  • Prefer scene-level exception statements over person-by-person ambiguous rules.

A practical approach is to secure recall with steps first, then improve precision by refining exceptions during operation.



Frame Mode Usage​

Stepwise mode supports both single-frame and multi-frame analysis.

  • Single-frame: single
  • Sequential flow: sequence
  • Time-window analysis: timeline

Example:

frame_mode:"sequence",
frame_mode     : "timeline"
frame_interval : 30

timeline is effective when temporal flow is important.



Technical Blog​

For deeper technical background, see From One-Shot Decisions to Two-Stage Reasoning.