Advanced Capstone Project
See what students create in the Applied AI Track
Cameron Y.
Age 16 | Applied AI Track Graduate
"Building my own neural network from scratch completely changed how I think about AI. This program didn't just teach me to use AI tools—it taught me to create them."
Project: PlantHealth AI
Advanced Capstone Project🌱 Project Overview
Cameron developed a sophisticated computer vision system that identifies plant diseases from photographs using a custom-trained convolutional neural network. The system analyzes leaf images, detects 15 different diseases across common houseplants and garden vegetables, and provides treatment recommendations. The full application includes a web interface where users can upload photos and receive instant diagnoses with 94% accuracy.
✨ Key Features
- Image Classification Model: Custom CNN trained on 8,000+ plant disease images using TensorFlow
- Real-time Analysis: Upload a photo and receive diagnosis in under 3 seconds
- Treatment Database: Comprehensive recommendations for organic and chemical treatments
- Disease Tracking: Users can track plant health over time with photo logs
- Confidence Scores: Each prediction includes probability percentages for transparency
- Mobile Responsive: Works seamlessly on phones for garden use
🛠️ Technologies & Algorithms Used
Core ML Techniques:
Convolutional Neural Networks (CNN), Data Augmentation, Transfer Learning (VGG16), Dropout Regularization, Adam Optimization, Cross-Entropy Loss Function
📸 Project Preview
🚀 Development Journey (6 Weeks)
Week 7: Research & Data Collection
Identified problem statement, researched existing solutions, collected and curated training dataset from multiple sources, set up development environment
Week 8: Model Architecture & Training
Designed CNN architecture, implemented data preprocessing pipeline, trained initial model, achieved 78% accuracy baseline
Week 9: Model Optimization
Applied transfer learning with pre-trained VGG16, implemented data augmentation, hyperparameter tuning, improved accuracy to 91%
Week 10: Backend Development
Built Flask API for model inference, created database for treatment recommendations, implemented image preprocessing pipeline
Week 11: Frontend & Deployment
Developed responsive web interface, integrated frontend with backend API, deployed to Azure cloud, conducted user testing
Week 12: Final Testing & Portfolio
Fixed edge cases, improved model to 94% accuracy, created comprehensive documentation, built portfolio website showcasing the project
🌟 Impact & Recognition
Cameron's PlantHealth AI has been used by over 200 local gardeners and farmers in his community. The project won 2nd place at his school's science fair and caught the attention of a local agricultural extension office. More importantly, Marcus used this portfolio project in his college applications, helping him secure acceptance to top computer science programs. The experience gave him the confidence to pursue AI engineering as a career path.
💡 Technical Highlights
- Implemented custom CNN with 5 convolutional layers, pooling, and fully connected layers
- Used data augmentation (rotation, flip, zoom) to expand training set from 2,000 to 8,000 images
- Applied transfer learning with VGG16 pre-trained weights, fine-tuning top layers
- Optimized inference speed to 2.8 seconds per image through model quantization
- Deployed using Docker containers on Azure App Service with auto-scaling
Ready to Build Your Own Advanced AI System?
Master machine learning, deep learning, and real-world AI development in 12 weeks!