Detecting Disease in Coffee Plans Using AI

Background

Ethiopia, the birthplace of coffee, is the largest coffee producer in Sub-Saharan Africa and ranks fifth globally, contributing 7-10% of the world’s coffee production. Coffee provides livelihoods for at least 15 million Ethiopians, with over 80% of coffee growers being small-scale farmers. However, major fungal diseases like coffee berry disease (CBD), coffee wilt disease (CWD), and coffee leaf rust (CLR) significantly reduce coffee production and consumption. The traditional method of disease surveillance—manual observation—is time-consuming, costly, and requires expertise, creating a pressing need for automated detection solutions.

Objective

• Analyze common coffee plant diseases.
• Source and preprocess training datasets.
• Develop and train a deep learning model for disease detection.
• Perform model inference on test data.
• Build a user-friendly mobile app for real-time disease identification.

Approach

The team tackled the challenge using advanced AI techniques:

1. Data Collection: Studied common diseases affecting Ethiopian coffee plants and sourced training datasets.
2. Data Preprocessing: Enhanced dataset quality and selected pre-trained models suitable for fine-tuning.
3. Model Development: Built a deep learning model leveraging image classification, segmentation, and object detection to identify disease types and assess severity.
4. Training and Testing: Trained the model with labeled data and validated its performance on unseen test data.
5. Mobile App Development: Designed an intuitive mobile application for farmers to diagnose diseases via image uploads, bringing AI solutions directly to end-users.

Tools and techniques included deep learning frameworks, image classification algorithms, and object detection methods.

Results and Impact

The project successfully developed a deep learning-based model capable of detecting CBD, CWD, and CLR in coffee plants with high accuracy. The mobile app empowered farmers with real-time disease identification tools, reducing reliance on expert observation and minimizing costs. This initiative enhanced coffee production efficiency, safeguarded livelihoods, and promoted sustainable agricultural practices in Ethiopia.

Future Implications

This approach can transform agricultural disease management across various crops and regions. Future developments could include integrating weather data for predictive analytics, expanding the model’s scope to other coffee-growing regions, and enabling multilingual support in the mobile app to cater to diverse user bases. The findings could inform agricultural policies and drive further research into AI-driven crop health management.