CODING PATH C

AI Pose Rescue Game

Train a Google Teachable Machine Pose Classification model and use body movements to control a rescue game character in MRT AI Studio.

🧠 Upper Primary – Year 12👤 Individual or Team of 2–5🤖 AI Pose Control

How to Use Google Teachable Machine 🧠

The purpose of this activity is to help students understand how to use Google Teachable Machine. Students will create and train their own pose model, upload the trained model, connect it to their game code, and use it to control one sprite or object in their game.

Your Mission 🎯

Train a Pose Classification model using Google Teachable Machine, then connect it to MRT AI Studio to create a rescue game controlled by your body movements.

For example: raise your arms to make the character jump over collapsed debris, lean left to dodge falling rocks, or hold a stop pose to halt before a danger zone.

💡 This is not about adding AI as decoration — AI must be the primary control method for your game character.

AI Requirements 🤖

Item	Requirement
AI Model	Google Teachable Machine — Pose Project only
Minimum Poses	At least 3 distinct poses (4+ recommended)
Game Platform	MRT AI Studio (block coding)
Not Accepted	Image Classification or Sound Classification as main control
Not Allowed	External ML frameworks (TensorFlow, PyTorch, etc.)

Pose Control Examples 🕹️

Pose	Game Action
Both arms raised	Jump or send rescue signal
Lean right	Move right
Lean left	Move left
Hand pushed forward	Stop
One hand raised	Request rescue
Crouching	Duck under obstacles

Project Examples 💡

Project	Description
AI Escape Game	Control a character with body movements to escape a collapsing highway zone
AI Rescue Signal Game	Recognise specific poses to send rescue signals from a disaster zone
AI Obstacle Dodge	Use poses to dodge falling rocks, cracks, and debris
AI Supply Delivery	Control a drone/character via body movements to deliver rescue supplies
AI Safety Training	An educational game that teaches correct rescue signal poses

Allowed Tools 🛠️

✅ MRT AI Studio✅ Google Teachable Machine (Pose only)

Not allowed as main control:

Image ClassificationSound ClassificationTensorFlow / PyTorchScratch (scratch.mit.edu)Unity / Python / JavaScript

Evaluation Criteria 📊

Criteria	Beginning (1)	Developing (2)	Accomplished (3)	Exemplary (4)	Weight
AI Model Quality	Model doesn’t recognise poses	Recognises 1–2 poses inconsistently	Recognises 3+ poses accurately	Stable, diverse training data, 4+ poses	×5 = 20
Game Design	No clear game goal	Basic concept, limited interaction	Clear mission, obstacles, scoring	Polished game with levels and progression	×5 = 20
Theme Connection	Not related to theme	Loosely related	Clear rescue scenario	Thoughtful, realistic rescue design	×5 = 20
Technical Integration	AI not connected	Partial connection, unstable	AI controls game actions reliably	Seamless AI-to-game integration	×5 = 20
Presentation	No explanation	Brief or unclear	Clear demo of AI + game	Excellent walkthrough of AI training + gameplay	×4 = 20

Total: 100 points

How to Get Started 🚀

All AI training guides, MRT AI Studio tutorials, Teachable Machine Pose setup instructions, project examples, and submission guidelines are provided through the MRT eLearning courses after registration.

Register and access the eLearning platform
Complete the AI Pose training course
Create your Pose model on Google Teachable Machine
Build your rescue game in MRT AI Studio
Connect your Pose model to the game
Test, improve, and refine
Record your presentation and demo video (maximum 5 minutes)
Submit via the Manager Dashboard

📌 Students do not need to choose their category during registration. Submit when your project is ready.

Submission Requirements 📦

A Presentation Video — maximum 5 minutes
A Teachable Machine model URL
A required Code File — SB3 only
An optional Project Note — two to five sentences if already prepared; it may be left blank

📹 Your video should show the AI pose training process AND the live gameplay demo. Voice explanation by the student is strongly recommended.

Upload the video and code file directly through the RCC Dashboard; Google Drive and participant YouTube links are not required. Explain the trained poses, controls, testing and learning in the video; a written report is not required.

Australian Curriculum Connections 🇦🇺

Curriculum Area	Connection
Digital Technologies	AI model training, data collection, input/output, algorithmic game control, digital solution creation
Mathematics	Data classification, accuracy analysis, patterns, conditional logic
Health & PE	Body awareness, movement, pose recognition, physical engagement
Critical & Creative Thinking	Designing AI-controlled solutions, evaluating model performance
Digital Literacy	Using AI tools, managing digital projects, understanding data privacy

Aligned with Australian Curriculum V9.0

“Train your AI. Move your body. Save lives through code.”