AIBox event verification software

A software capability within AIBox: CV screens candidate events, then a large model checks whether they are valid.

Two-stage event review

The CV module performs initial screening. A large model then reviews candidate events. Each stage is evaluated in the context of the actual task.

Walk through verification
  1. 01
    Visual input

    Event material to review

  2. 02
    CV screening

    Screen for candidate events

  3. 03
    Large-model review

    Assess event validity

  4. 04
    Review result

    Feed operational review and training data

Workflow illustration. Inputs and supported events are confirmed during evaluation.

What to measure during evaluation

Compare screening with the full workflow on the same reference-labelled samples.

01

Are valid events retained?

Measure valid-event detection and valid events filtered out, alongside the amount of filtering.

02

How are invalid events handled?

Review invalid events before and after verification, including cases requiring human review.

03

What does operation require?

Check input interfaces, latency, resource use and integration with the existing workflow.

Common questions

Why use two stages?

CV performs initial screening, while the large model focuses on candidate review. Evaluation with actual data determines suitability.

Can verification recover events missed by screening?

Events that never enter the candidate set do not reach the second stage. Screening misses must be measured separately.

How are deployment and compatibility scoped?

Start with the inputs, existing devices and runtime environment, then confirm supported events, interfaces, models and service scope.

Product enquiries

Tell us which product you need, your current equipment and the problem you want to solve.