AI-Powered Early Detection of Crop Diseases in Kenyan Smallholder Farms

Challenge Background

Kenya's agricultural sector is dominated by smallholder farmers who produce a significant portion of the country's food. Maize, beans, and cassava are staple crops, but crop diseases frequently threaten food security and farmers' livelihoods.

The Problem

Smallholder farmers in Kenya often lack access to timely and accurate crop disease diagnosis, leading to significant crop losses and reduced food security. Early detection and treatment of crop diseases are crucial but challenging due to limited agricultural extension services.

Goal of the Project

1. Detection Accuracy:

• Achieve 95% accuracy in identifying common diseases for maize, beans, and cassava.
• Reach 90% accuracy in distinguishing between different stages of disease progression.

2. Language Support:

• Provide fully functional disease identification and recommendations in both Swahili and English.
• Achieve 98% accuracy in translating technical agricultural terms between the two languages.

3. User Adoption:

• Onboard 100,000 smallholder farmers across Kenya within the first year.
• Maintain an 80% monthly active user rate after six months.

4. Disease Coverage:

• Include at least 20 common diseases for each crop (maize, beans, and cassava) in the initial release.
• Expand to cover at least 5 additional crops within two years.

5. Impact on Crop Yield:

• Contribute to a 20% reduction in crop losses due to diseases among app users within the first growing season.
• Help increase overall crop yields by 15% for active users within two growing seasons.

6. Offline Functionality:

• Ensure core disease detection features work offline for areas with limited connectivity.
• Achieve 90% accuracy in offline mode compared to online performance.

7. System Performance:

• Process and provide results for uploaded images within 5 seconds on average.
• Handle up to 50,000 simultaneous users without performance degradation.

8. Community Engagement:

• Facilitate the submission of at least 500,000 new, labeled images from users for continuous model improvement.
• Achieve a user satisfaction rate of 85% based on in-app feedback.

9. Integration with Extension Services:

• Partner with at least 5 regional agricultural extension offices for app distribution and support.
• Reduce the workload on extension officers for routine disease identification by 40%.

10. Economic Impact:

• Contribute to a 10% increase in income for smallholder farmers using the app regularly.
• Demonstrate a positive return on investment for farmers within one growing season.

Project Timeline

What you'll learn

1. Machine Learning for Agriculture:

• Gain proficiency in developing and fine-tuning computer vision models for plant disease detection.
• Learn techniques for handling imbalanced datasets common in agricultural applications.

2. Mobile Development for Low-Resource Settings:

• Develop skills in creating efficient, offline-capable mobile applications.
• Learn strategies for optimizing AI models to run on low-end smartphones.

3. Agricultural Science:

• Gain knowledge about common crop diseases, their progression, and management strategies.
• Understand the specific agricultural challenges faced by smallholder farmers in Kenya.

4. Natural Language Processing:

• Develop skills in creating multilingual systems, particularly for languages with limited digital resources.
• Learn techniques for translating and presenting technical agricultural information in local languages.

5. Data Collection and Annotation:

• Gain experience in designing effective protocols for crowdsourced data collection.
• Learn best practices for annotating agricultural images for machine learning.

6. User Experience Design for Diverse Populations:

• Develop skills in creating intuitive interfaces for users with varying levels of technological literacy.
• Learn to design effective visual communication systems for conveying complex agricultural information.

7. Community-Centered Design:

• Gain experience in participatory design methods involving smallholder farmers.
• Learn techniques for incorporating indigenous knowledge into AI systems.

9. Impact Evaluation:

•  Develop skills in designing and conducting studies to measure the economic and social impact of agricultural AI.
• Learn to use both quantitative and qualitative methods to assess technology adoption and effectiveness.

10. Scalable Cloud Architecture:

• Gain experience in designing cloud-based systems that can handle sporadic connectivity and data sync.
• Learn to create scalable architectures that can expand to cover more crops and regions.

11. Interdisciplinary Collaboration:

• Develop skills in working effectively with agronomists, extension officers, and rural development experts.
• Learn to communicate complex technical concepts to non-technical stakeholders.

12. Continuous Learning Systems:

• Gain experience in developing AI models that improve over time with user-generated data.
• Learn techniques for active learning and semi-supervised learning in agricultural contexts.

13. Policy and Regulatory Compliance:

• Understand the regulatory landscape for agtech solutions in Kenya and East Africa.
• Learn about data protection and sovereignty issues in agricultural AI applications.