Build Your Own AI
Lesson 12: Hands-On AI Project Development

Evolve AI Institute | Free AI Education for All

What You'll Learn Today
  • Create a functional AI application from scratch
  • Apply the complete AI development lifecycle
  • Collect and prepare training data effectively
  • Test, evaluate, and improve your AI model
  • Present your AI project professionally
The AI Development Process
Plan → Collect Data → Train → Test → Iterate → Present

1. Plan Your Project

Define the problem, choose your platform, identify categories

2. Collect Training Data

Gather 50+ diverse examples per category

3. Train Your Model

Let the AI learn from your examples

4. Test & Evaluate

Try your AI in different conditions, measure accuracy

5. Iterate & Improve

Add more data, retrain, test again

6. Present Your Work

Demonstrate and explain your AI project

Choose Your Platform
🎓

Teachable Machine

Quick AI experiments

Image, sound, & pose recognition

Easiest to learn!

📱

MIT App Inventor

Build real mobile apps

AI-powered features

Create apps you can share!

🐱

Scratch with ML

Interactive games & stories

Creative AI projects

Most creative freedom!

🎓 Teachable Machine Quick Start

Step 1: Access the Platform

Go to teachablemachine.withgoogle.com → Click "Get Started"

Step 2: Choose Project Type

Select Image, Audio, or Pose Project based on your idea

Step 3: Add Classes & Collect Samples

Create 3-5 categories, record 50+ samples per class with variety

Step 4: Train & Test

Click "Train Model" → Wait 1-3 minutes → Test in Preview panel

📱 MIT App Inventor Quick Start

Step 1: Create Account & Project

Go to appinventor.mit.edu → Sign in → Create New Project

Step 2: Design Interface (Designer View)

Add Button, Camera, Image, Label components

Step 3: Add AI Extension

Import "Personal Image Classifier" or "ImageBot" extension

Step 4: Program Logic (Blocks View)

Connect blocks: Button Click → Take Picture → Classify → Display Result

🐱 Scratch with ML Quick Start

Step 1: Access Scratch

Go to scratch.mit.edu → Sign in → Click "Create"

Step 2: Add ML Extension

Click "Add Extension" → Select "Video Sensing" or use ML for Kids

Step 3: Train Model (ML for Kids)

Go to machinelearningforkids.co.uk → Train custom model → Import to Scratch

Step 4: Program & Test

Use ML blocks with game logic → Test → Share project online

Training Data: The Key to Success
50+

Samples per category (minimum)

✓ Good Training Data

  • Diverse angles & positions
  • Different lighting conditions
  • Varied backgrounds
  • Multiple distances/sizes
  • Balanced across categories

✗ Poor Training Data

  • All identical conditions
  • Only one angle/position
  • Same background every time
  • Blurry or low quality
  • Unbalanced categories
Testing Your AI Model

Run These Tests:

  • Test with samples similar to training data
  • Test with different lighting & angles
  • Test edge cases & ambiguous inputs
  • Test with items not in any category
  • Have classmates test your AI
Target Accuracy: Aim for 70%+ accuracy. Professional AI systems aren't perfect, and neither is yours - that's okay! Focus on steady improvement through iteration.
The Power of Iteration
Train → Test → Improve → Repeat

If Accuracy is Low:

→ Add more diverse training samples

→ Check if categories are too similar

→ Remove poor quality samples

→ Simplify your categories

If Specific Classes Confuse:

→ Add contrasting examples

→ Increase variety in those classes

→ Check background interference

→ Test in target environment

Remember: Even professional AI developers iterate multiple times. It's a normal and important part of the process!
Common Mistakes to Avoid

❌ Not Enough Training Data

Solution: Collect minimum 50 samples per category

❌ Identical Training Samples

Solution: Vary angles, lighting, backgrounds, and positions

❌ Unbalanced Categories

Solution: Ensure roughly equal samples for each category

❌ Testing Only in One Condition

Solution: Test in various lighting, angles, and environments

❌ Giving Up After First Try

Solution: Embrace iteration - improvement takes multiple attempts!

AI Project Ideas

🖼️ Image Recognition

  • Recycling sorter
  • Plant identifier
  • Emotion detector
  • Food classifier
  • Animal recognizer

🎵 Sound Recognition

  • Musical instrument classifier
  • Animal sound detector
  • Voice command system
  • Environment sound monitor

🤸 Pose Detection

  • Yoga pose checker
  • Hand gesture controller
  • Sign language interpreter
  • Exercise counter
  • Dance game

📱 Mobile Apps

  • Study helper with image scanning
  • Accessibility aid
  • Shopping assistant
  • Language learning tool
Today's Timeline

⏰ 10 min - Introduction & Planning

Review project goals, choose platform, complete planning worksheet

⏰ 15 min - Platform Tutorial

Learn your chosen platform, understand interface and features

⏰ 20 min - Data Collection

Gather diverse training samples for your categories

⏰ 20 min - Training & Testing

Train model, run initial tests, evaluate performance

⏰ 20 min - Refinement

Improve model based on testing, add features, polish interface

⏰ 15 min - Showcase & Reflection

Present projects, gather feedback, reflect on learning

What Does Success Look Like?

You'll Demonstrate Mastery When You:

  • Create a functional AI with 70%+ accuracy
  • Collect 50+ diverse samples per category
  • Complete at least one iteration cycle
  • Clearly explain how your AI works
  • Identify limitations and areas for improvement
  • Successfully demonstrate your AI live
Tips for Success

💡 Start Simple

Begin with 3-4 categories, add more later if time allows

💡 Test Early & Often

Don't wait until the end to test - catch problems early

💡 Ask for Help

Your teacher and classmates are resources - don't struggle alone

💡 Save Your Work

Regularly save progress to avoid losing your project

💡 Have Fun!

This is your chance to be creative - enjoy the process of creation

Quick Troubleshooting Guide

🔧 Camera/Mic Not Working

Check browser permissions, try different browser

🔧 Low Accuracy

Add more diverse data, simplify categories

🔧 Training Takes Forever

Reduce sample count or image quality

🔧 Platform is Slow

Close other tabs, clear cache, restart browser

🔧 Can't Export Model

Check internet connection, try different format

🔧 Project Too Ambitious

Simplify scope, focus on core functionality

Presenting Your AI Project

Include in Your Presentation:

  • Project title and purpose (what problem does it solve?)
  • How it works (what type of AI, what categories)
  • Live demonstration with real input
  • Challenges faced and how you solved them
  • Accuracy results and testing data
  • Future improvements you'd like to make
Presentation Tip: Practice your demo beforehand! Make sure you have good lighting, clear audio, and test objects ready. Prepare for the possibility that something might not work perfectly - explain how you'd fix it.
Beyond Coding: What You're Really Learning

Technical Skills

  • Machine learning concepts
  • Data collection & preparation
  • Model training & evaluation
  • Testing & debugging
  • User interface design

Life Skills

  • Problem solving
  • Persistence & iteration
  • Critical thinking
  • Project management
  • Public speaking

You're not just building an AI project - you're developing skills that will serve you throughout your career!

Resources & Support

📚 Documentation

  • Project Planning Worksheet
  • Development Guide (all platforms)
  • Testing Checklist

🌐 Online Help

  • Platform tutorials
  • Community forums
  • Video guides

👥 Support

  • Your teacher
  • Classmates
  • Platform experts
Remember: Getting stuck is part of learning! Don't hesitate to ask for help when you need it.
Ready to Build Your AI?
Let's Get Started!

✓ Choose your platform

✓ Plan your project

✓ Start building!

You're about to become an AI creator!

1 / 20