PICTURE THIS · DAY 4 OF 5
Train It + Connect It
Design a movement-based classifier, check who it works for, then connect its predictions to code.
Topic: Model design and human-controlled AI actions
Objective: I can train two or three distinct classes and connect each prediction to a visible action in RAISE Playground.
Demonstration of learning: My partner can trigger the intended responses, and I can name one change that improved reliability.
Plan before you train
- Choose two or three actions that look different on camera: lean left/right, thumbs up/down, hands high/low, or another teacher-approved set.
- Decide what the sprite should do for each class.
- Complete the planning sheet with your partner.
Train + bias check
- Build the model in Teachable Machine with at least 30 varied images per class.
- Trade seats. Test a different person, distance, angle, lighting condition, and background.
- Add the examples your model was missing and retrain.
- Export the shareable model link and paste it in your Canvas submission comment.
Connect the prediction to code
- Open the RAISE Playground template.
- Load your Teachable Machine model link.
- Match every class name exactly.
- Give each class a different visible response. Get at least one response working before the bell.
Leave-up success criteria
- Two or three clearly different classes.
- At least 30 varied examples per class.
- A partner completed the bias check.
- Model link saved.
- At least one Playground response works.
Turn in: Upload the completed planning sheet as a PDF, Word file, photo, or scan. Paste the shareable model URL in your Canvas submission comment.
Help stem: My model detects ______, but the sprite does not ______. I already checked ______.
RAISE Playground was created by the MIT RAISE Initiative and the Personal Robots Group at the MIT Media Lab.