Frontend Homepage
Identify Disease Page
Identify Disease Page (With return)
Get AI Advice
Get AI Advice Page (With returned advice)
Get AI Advice Page (Renaming advice)
Get AI Advice Page (With saved advice)
Alert Map (Frontend Mock)

About CropCare - TuringToast

Inspiration

Crop diseases can cause devastating losses for farmers, especially when early symptoms go unnoticed. We were inspired to create CropCare to empower farmers with accessible, AI-driven tools that help them detect diseases early and take informed actions — minimizing crop damage and safeguarding livelihoods.

What it does

CropCare helps farmers identify crop diseases through a simple mobile app.
Users can upload a photo of an affected plant leaf, and the system:

Uses a computer vision classifier to predict the likely disease.
Retrieves targeted information from an AI-powered knowledge base using RAG.
Provides actionable insights and management tips to help farmers respond quickly.

How we built it

Computer Vision Classifier: An end-to-end deep learning pipeline for crop health detection based on MobileNet_v2, from data preprocessing and model training to real-world inference via a deployable API interface.
RAG System: A retrieval-augmented generation system fetches detailed disease information and generates user-friendly advice.
Mobile Frontend: A mobile-optimized web application allows farmers to easily upload images and view results.
Cloud Deployment:
- The FAISS vector database and knowledge documents are hosted on Azure Blob Storage for scalable and efficient access.
- Azure OpenAI powers the language generation component of the RAG system.
- Both the computer vision and RAG APIs are deployed on a cloud-based virtual machine to ensure flexibility, control, and performance.

Challenges we ran into

Data Quality: It was challenging to source high-quality, diverse images for training the CV model.
RAG Fine-tuning: Ensuring retrieved information was accurate, relevant, and concise required several iterations.
System Integration: Connecting the classifier, RAG system, and mobile frontend seamlessly demanded careful API and pipeline design.
Mobile UX: Optimizing user experience across different devices and network conditions was more complex than expected.

Accomplishments that we're proud of

Successfully integrated computer vision and knowledge retrieval into a smooth user flow.
Built a working mobile-first web app prototype within a tight timeline.
Created a scalable architecture that can be extended to support more crops and diseases in the future.

What we learned

How to apply computer vision and RAG technologies to solve real-world agricultural problems.
Best practices for building mobile-first web apps that interact with AI services.
The importance of iteration, especially when balancing model performance, retrieval quality, and frontend usability.

What's next for CropCare

Expanding the crop and disease database to cover a wider range of plants.
Improving the computer vision model with more training data and advanced architectures.
Continuously expanding and curating the knowledge base to ensure accurate, reliable, and professional responses for farmers.
Adding location-based disease alerts to help farmers anticipate regional outbreaks.
Exploring partnerships with agricultural organisations to bring CropCare to real-world users.