How TeaPulse Works
From satellite land selection to daily soil analysis, guided tasks, and yield predictions — TeaPulse catches problems before they ever reach the leaf.
Tech Stack
Platform Architecture
Interactive dashboards, harvest input studios, and real-time image preview for AI diagnosis.
Manages grade-wise harvest logs (Grade A & B) with secure Row Level Security (RLS).
MobileNetV2 transfer learning model classifying 8 tea leaf conditions instantly via API.
Services & Core Modules
Key Operational Flows
Manager uploads a tea leaf image via the Next.js frontend diagnostic interface.
Frontend sends the image file securely via FormData to the Python Flask AI API (:5000).
Flask backend preprocesses the image using NumPy and Pillow (PIL) to match MobileNetV2 requirements (224x224).
The TensorFlow/Keras CNN model evaluates the image across 8 classes, returning the detected disease and confidence score.
Frontend dynamically renders the result card along with specific agricultural remedy instructions.
Engineering & AI Highlights
Ready to explore the code?
View on GitHub (HirushaDulshaan/TeaPulse-Smart-Estate-Platform) ↗