Can you catch what the cameras see?

Human mode: select a seat in the room or seat grid, choose an activity, then flag it while it happens. Correct flags earn 10 points, false flags lose 4, and missed events lose 2. The answers stay hidden until the round ends. You can pause to inspect the room.

AI mode: two locally trained neural networks combine their class probabilities from noisy synthetic observations across the enabled cameras. They cannot read the hidden answer key. Change the coverage and concealment to test their limits. Rule, untrained and random controllers are available for comparison.

Eight camera positions include two low angles. Select a camera to inspect its coverage, or use the combined view. These are schematics, not video feeds. Phones, gestures and transfers are synthetic. The NN does not analyse pixels, audio or real students. Browser model probabilities are not real-world confidence or evidence of misconduct.