Our solution is smart and intelligent to provides timely, context-specific information that enables smallholder farmers and those who support them to identify, suggest treatment, track and share information about incidences of pests and diseases on African farms and gardens.
It uses machine learning on Android to help farmers fight pests and disease affecting their crops. It is designed to work both on and offline.
Pests and diseases on African farms not only cause a threat to food security but also to the economic prosperity of farmers. According to the report commissioned by the Department for International Development(DFID) in September, 2017, the arrival of fall armyworm (FAW) in Africa had the potential to cause maize yield losses in arange from 8.3 to 20.6 million tonnes per annum, in the absence of any control methods, in just 12 maize-producing countries. This represents a range of 21%-53% of the annual averaged production of maize over a three year period in these countries. The value of these losses was estimated between $2,481m - $6,187m. The same worm can affect 80 plant species.
Two thirds of the population in poor countries depend on agriculture for survival. 30-40 % of agriculture yields are lost due to pest, diseases and post-harvest loss every year causing a big threat to food security to the population. This idea when implemented will lead to an improvement in yields due to early detection and treatment of crops from pests and diseases. The information collected about the detected and treated pests, diseases and crops with farmers’ consent will be data mined and used for future predictions of when outbreaks are likely to occur, which areas are more susceptible, etc. This will lead to minimized or elimination of these pests and diseases hence no hunger. Time to know and acquire the pesticides and advice will be reduced because the solution will not only suggest the best pesticide and action to take but also the nearest possible store to purchase or obtain the pesticide.
Our solution uses TensorFlow, an open source software library for high performance numerical computation with a flexible architecture that allows easy deployment of computation across a variety of platforms. TensorFlow comes with strong support for machine learning and deep learning with a flexible numerical computation core that is used across many other scientific domains.
We collected image data of pests and crops featuring signs and symptoms of particular disease or pest infestation. Data collected was carefully sorted and processed. The well sorted and processed data was used in training a deep learning model. With a trained model available, it integrated in a mobile application that is used on low end smartphones to deliver the functions. The mobile application uses the vast capabilities and items onboard like sensors and camera to deliver the targeted experience.
A number of challenges were and still are affecting this agricultural innovation, below are the salient ones. Lack of resources to collect enough training data and high spec equipment to use the collected data to train a machine learning model. Many smallholder farmers do not have smartphones to use the solution The team did not have experts like agronomists, pathologists at the time, we have just gotten And now formalizing a co-working relationship with Buginyanya Zonal Agricultural Research and Development Institute (BugiZARDI). Lack of access to good equipment during development is another constraints