AGROGUARD

• 

  demo page of showing disease detected

//Inspiration

Farmers often detect crop diseases only after visible damage has already spread across the field. Manual inspection is time-consuming, expert advice is not always accessible, and overuse of pesticides increases cost and environmental harm. We were inspired to build AGROGUARD to provide early, automated, and affordable crop disease detection directly from the field using IoT and AI, without depending on smartphones or constant expert intervention.

// What it does

AGROGUARD is an IoT-enabled AI crop monitoring system that:

Automatically captures crop leaf images using ESP32-CAM

Collects temperature, humidity, and soil moisture data

Detects diseases in rice and sugarcane crops using a deep learning model

Provides crop-specific treatment and prevention recommendations

Displays real-time crop health insights through a web-based dashboard

Alerts users when disease risk or environmental stress is detected

//How we built it

Hardware Layer:

ESP32-CAM for image capture

ESP32 with DHT and soil moisture sensors for environmental data

Communication Layer:

Wi-Fi-based data transmission to the backend

AI Layer:

Trained a deep learning model on labeled crop disease images

Optimized the model for CPU-based inference

Backend Layer:

REST API for receiving sensor data and images

Disease prediction and recommendation engine

Frontend Layer:

Web application displaying disease status, confidence score, and actionable recommendations

//Challenges we ran into

Handling image quality issues under varying outdoor lighting conditions

Managing limited processing power of edge devices

Ensuring reliable connectivity in agricultural environments

Balancing model accuracy with real-time performance

Designing recommendations that are simple and farmer-friendly

//Accomplishments that we're proud of

Built a fully automated system without using farmer smartphones

Successfully integrated IoT + AI + recommendation system

Achieved early-stage disease detection in real field conditions

Designed a low-cost and scalable architecture suitable for small farms

Created a solution with real social and agricultural impact

// What we learned

Real-world AI systems require hardware-software co-design

Data quality is as important as model complexity

Agriculture needs simple, actionable outputs, not just predictions

Edge and cloud systems must work together efficiently

Technology adoption depends on usability and trust

// What's next for AGROGUARD

Add support for more crops and diseases

Integrate weather forecasting for predictive disease alerts

Enable offline edge inference for low-connectivity areas

Introduce multilingual voice-based recommendations

Deploy pilot projects in real farming communities

Collaborate with agricultural departments and NGOs

Built With

• bhuvan
• fastapi
• ml
• model
• python
• react