← All materials

Mission 12 kit · screen and print

Voice and vision capstone

A complete AI game turns live sensor inputs into uncertain model outputs, then applies thresholds and an exact fusion rule before acting or abstaining.

1OpenKeep this page beside the mission.
2UseTap, point, tally, or move pieces on screen.
3Print if usefulThe same content appears in print mode.
Piece 1 · checklist · stars pattern

Finale setup checklist

How to use it: Confirm the external tools, prop zone, unchanged Audio Model v1, thresholds, alternative control, and adult spotter before the live round.

01

Adult spotter ready

02

Prop-only webcam zone

03

Audio Model v1 unchanged

04

Image input ready

05

AND rule visible

06

Alternative control ready

07

Stop permissions after play

Piece 2 · log · grid pattern

Live FusionTrace

How to use it: Write the observed live image and audio results, thresholds, fusion rule, and final state while the adult operates the external tabs.

01

Live image label and score

02

Live audio label and score

03

Two thresholds

04

AND rule

05

ACT or ASK AGAIN

06

Human override

Piece 3 · cards · rings pattern

ACT and ASK AGAIN proof card

How to use it: Complete one safe ACT case and one safe ASK AGAIN case. Never force an uncertain input.

01

ACT case: both pass

02

ASK AGAIN case: one or both fail

03

HUMAN DECIDES remains available

Piece 4 · log · stripes pattern

Final dev log

How to use it: Explain the complete perception-to-action chain and record only the adult's manual confirmation.

01

What each model sensed

02

What the AND rule decided

03

Why ASK AGAIN matters

04

When HUMAN DECIDES

05

Adult manually confirmed

Safe mission inputs

Use the listed props and game signals.

Aim the sensors at safe props and game commands in the supervised Google tool, then turn them off when play ends.