crophelthh

AI Smart Crop Health Monitoring Robot - Model Card

Project Overview

This repository contains machine learning models trained for the AI Smart Crop Health Monitoring Robot, developed as a 1st-year Semester 1 mini IoT project for the Fundamentals of Computing (IT1140 / IT1040) module at the Sri Lanka Institute of Information Technology (SLIIT).

The autonomous robotic system is designed to monitor crop health, collect environmental and soil data, and detect plant diseases using advanced computer vision to support precision agriculture.

---

Project Team (Group 01 - Kurunegala)

• IT26101404 - P.G. Hansaka Rasanjana
• IT26101824 - T.D. Avishka Dewinda
• IT26101774 - P.G. Amalka Sandanayani
• IT26102072 - N.D. Maddumage
• IT26100283 - M.K.M. Raamy Khaleel

---

Model Details

• Base Backbone: EfficientNetB0 (used for feature extraction and image classification)
• Task: Plant Leaf Disease Detection and Classification
• Target Classes: Identifies common crop issues such as leaf spot, powdery mildew, rust, blight, yellow mosaic, and nutrient deficiencies.
• Frameworks: Python, TensorFlow / Keras, OpenCV
• Hardware Target: ESP32-S3 N16R8 Microcontroller deployment / Server-side inference

---

Datasets

The models in this repository were trained and fine-tuned using custom agricultural datasets hosted on Hugging Face:

1. hansaka01/crophelth - Primary training dataset.
2. hansaka01/crophelth_finetune_2nd_stage - Second-stage fine-tuning dataset for enhanced model accuracy.

---

System Architecture Highlights

• Perception & Hardware Layer: Powered by an ESP32-S3 microcontroller, 4WD robotic chassis, AI Camera, soil moisture, temperature/humidity (DHT22), rain, and obstacle detection sensors.
• Communication Layer: Real-time telemetry via WebSockets (Socket.io) and HTTP REST APIs over Wi-Fi.
• Backend Processing: Node.js and TypeScript server handling telemetry rules and automated alerts.
• Client Layer: React / Vite web dashboard and mobile interface for remote tracking.

---

Usage

These model weights can be utilized in Python environments for plant disease image classification tasks or integrated into your project pipeline.