This page shows the real time results from a survey system developed at Makerere University, Uganda, which uses camera-phone input to provide timely data on the health of staple crops. Survey workers (with GPS enabled devices) and agricultural extension workers or farmers (with basic camera phones) can provide data in the form of images taken of the leaves of their crops. Applying computer vision techniques to large sets of such uploaded images, we can automatically classify the state of health of plants, and then map the extent of the disease in a district or country. In this way, more data can be collected, more rapidly and at lower cost, than is possible with traditional survey methods.
We concentrate here on cassava, which is the third largest source of carbohydrate for human consumption in the world, and provides more food calories per unit of land than any other staple crop.
More information on how the system works.
Publications and technical reports
J. Quinn, K. Leyton-Brown, E. Mwebaze. Modeling and Monitoring Crop Disease in Developing Countries. Conference of the Association for the Advancement of Artificial Intelligence (AAAI), Computational Sustainability and AI Track, San Francisco, 2011. [pdf]
J.R. Aduwo, E. Mwebaze and J.A. Quinn. Automated Vision-Based Diagnosis of Cassava Mosaic Disease, Workshop on Data Mining in Agriculture, Berlin, 2010. [pdf]
One page summary of the monitoring system [pdf]
Guy Acellam, Makerere University AI-DEV
Jennifer Aduwo, Makerere University AI-DEV
Titus Alicai, NaCRRI
Kevin Leyton-Brown, University of British Columbia
Ernest Mwebaze, Makerere University AI-DEV
John Quinn, Makerere University AI-DEV
Feedback/enquiries: please contact jquinn@*nospam*cit.ac.ug
This video shows examples of our computer vision software running on a low cost Android phone (Huawei Ideos, available in Kenya for $85 as of end 2011). No network connection is needed for basic diagnosis, though with connection to a server more is possible such as counting parasites on the leaf, and diagnosis with higher accuracy.