DiseaseDetectPro: Agentic Crop Vision AI

Inspiration

Crop diseases threaten agricultural productivity, farmers' livelihoods, and food security. In many agricultural communities, access to timely disease diagnosis and expert guidance can be limited, making early detection and reliable information particularly important.

As a computer science lecturer, researcher, and AI developer, I have explored how artificial intelligence can address agricultural challenges. My previous research on orange and potato disease detection demonstrated the potential of deep learning and explainable AI to identify crop diseases and support informed agricultural decisions.

Building on this experience, I developed the concept of DiseaseDetectPro VisionAgent, an intelligent vision system that combines OpenCV 5, AWS, agentic AI, and explainability to move beyond conventional image classification.

My vision is to create an AI system that not only identifies potential crop diseases but also helps users understand predictions and access useful agricultural information.

What it does

DiseaseDetectPro VisionAgent aims to transform crop images into meaningful information through an AI-powered computer vision workflow.

Its core objectives are to:

Detect crop diseases: Analyze crop images using AI-based vision models.

Explain predictions: Help users understand model outputs through appropriate visual explanations and supporting information.

Enable agentic vision: Coordinate image analysis and interpretation through an AI agent, according to the capabilities implemented.

Support agricultural decisions: Provide relevant disease information and, where implemented and validated, crop management recommendations.

Leverage AWS: Explore cloud-based AI integration and scalable application deployment.

The goal is to make AI-assisted crop disease information more accessible to farmers and agricultural professionals.

How we built it

The project brings together computer vision, AI-based disease detection, explainable AI, agentic workflows, and cloud technologies.

OpenCV 5: Provides computer vision and image-processing capabilities within the proposed crop image analysis pipeline.

AI-based disease detection: Uses an AI vision model to identify disease categories supported by the model. My previous research in crop disease detection provides a foundation for this work.

Explainable AI: Builds on my experience with Grad-CAM and Grad-CAM++ to explore how visual evidence can help users interpret model predictions.

Agentic vision: Explores how an AI agent can coordinate image analysis, interpret available results, and organize relevant information into a useful response.

AWS integration: Supports the project's cloud-based AI architecture and potential deployment and scalability.

The overall design connects image input, visual analysis, disease prediction, explanation, and the presentation of useful information. The specific functionality depends on the components implemented and tested in the application.

Challenges we ran into

One major challenge is the variability of crop images. Lighting, background complexity, leaf orientation, image quality, and disease severity can influence model performance. Real-world validation is essential to determine how reliably a system performs outside controlled datasets.

Another challenge is integrating computer vision, model inference, explainability, agentic workflows, and cloud services into one coherent application.

Explainability also requires care: visualizations must not be presented as definitive proof that a disease diagnosis is correct. Similarly, an AI agent must avoid unsupported conclusions and unverified treatment recommendations.

Cloud integration introduces additional considerations, including configuration, latency, reliability, deployment complexity, and operating costs.

These challenges reinforce the importance of modular design, systematic testing, transparent reporting, and continuous improvement.

Accomplishments that we're proud of

i am proud to develop DiseaseDetectPro VisionAgent: An OpenCV 5 and AWS-Powered Agentic Vision System for Explainable Crop Disease Detection, an initiative that explores how modern AI technologies can address real-world agricultural challenges.

my key accomplishments include:

• Combining computer vision with agricultural AI: Building on the potential of OpenCV 5 and AI-based vision models to support crop disease detection and analysis.

• Advancing explainable AI: Making transparency a central design goal so users can better understand AI predictions rather than relying on unexplained classifications.

• Exploring agentic vision: Developing an approach that aims to connect image analysis, interpretation, and agricultural information within a unified AI workflow.

• Integrating cloud-powered AI: Exploring how AWS can support AI integration and the development of scalable agricultural applications.

• Building on previous research: Applying knowledge gained from my research on orange and potato disease detection, including deep learning, transfer learning, and explainable AI techniques such as Grad-CAM and Grad-CAM++.

• Focusing on real-world impact: Addressing the need for accessible crop disease information, particularly for agricultural communities where timely expert assistance may be limited.

Beyond the technology, we are proud of the project's broader vision: moving from simply identifying potential crop diseases toward helping people understand AI predictions and make better-informed agricultural decisions.

my greatest accomplishment is bringing together agricultural research, computer vision, explainable AI, and agentic system design around a meaningful real-world problem. This project represents an important step toward developing more transparent, accessible, and practical AI solutions for agriculture.

What we learned

This project has strengthened my understanding of integrating computer vision, explainable AI, agentic workflows, and cloud technologies into a practical application.

I learned that building a useful AI system involves more than achieving high classification accuracy. Models must also be evaluated under realistic conditions, predictions should be communicated responsibly, and explanations should help users understand rather than blindly trust AI outputs.

I also gained a deeper appreciation of the challenges of integrating different technologies, designing reliable workflows, and considering scalability, usability, and responsible AI deployment.

My previous agricultural AI research helped me approach this project with a clearer understanding of disease classification, transfer learning, and explainability.

What's next for DiseaseDetectPro: Agentic Crop Vision AI

My goal is to develop DiseaseDetectPro VisionAgent into a practical AI-assisted agricultural tool that helps farmers and agricultural professionals access crop disease information more efficiently.

Future development could include additional crops and disease categories, validation using locally collected agricultural images, improved explanations, local-language support, validated crop management guidance, and mobile access.

For agricultural communities in Ethiopia and beyond, these capabilities could contribute to more accessible disease information and better-informed agricultural decisions.

The system's effectiveness and impact will need to be established through implementation, rigorous testing, and feedback from intended users.

Built With

• agentic
• ai
• amazon-web-services
• cloud
• computer-vision
• computing
• deep-learn
• explainable-ai
• flask
• grad-cam
• grad-cam++
• image-process
• machine-learning
• opencv
• python
• transfer-learning