Inspiration

The Farm AI App is inspired by the needs of smallholder farmers. These farmers in rural communities mostly lack access to reliable information and real-time help with regards to pest control and disease management. However, with our mobile solutions, smallholder farmers can now have easy access to expert-level knowledge with a tap of a button – with or without internet connectivity. Our machine learning platforms enable farmers to detect pest infestations at an early stage and take actions to avoid spread and further damage to their crops. This means farmers will spend little or no money on pesticides and have increased yield, enabling them to support themselves and their families beyond the farming season.

What it does

FarmAi is a Zimbabwean-based small and medium enterprise (SME) that has developed a Smartphone application that utilizes artificial intelligence (AI), machine learning (ML), and data analytics to predict and detect crop diseases and infestations and offers recommended solutions, based on scientific knowledge, in real-time. The application was designed to be used by rural farmers in Zimbabwe and operates without an internet connection. The application has allowed farmers to reduce their expenses related to purchasing crop pesticides, and increase their crop productivity and harvest by up to 50 percent, thereby increasing their profits and disposable income. Our mission is to minimize hunger and poverty in Africa by building tech solutions to help farmers increase yield and prevent losses.

The Farm AI App uses an innovative user interface design. The interface allows you to take a photo of a crop, then the app analyses the crop and detects the disease (if any). The predicted disease is shown as a number (Example: 1, 2, 3, 4…). The farmer can then tap on the number and get further insights. The insights are delivered in a local dialect (vernacular) to enable farmers to take actions that are sustainable to prevent losses. Recommendations are in the form of an animated video that explains exactly how the farmer should handle the infestation.

Our applications are trained with images of crops both healthy and infested (example images). The machine learning model is built to learn from these images, learn the patterns and features of diseases. With an average accuracy of 93.3%, the applications are then able to predict whether a given image is healthy or infested. Due to the complexity of our algorithms, the machine learning model can extract very fine details which enable it to detect diseases at an early stage even before the human eye Notices it. The model is “frozen” and embedded in the application and runs on the device locally without the need for internet connectivity to make an inference

How we built it

The FarmAi app is currently in its developmental stage, where we will use python language and we will create an end-to-end Android application with TFLite that will then be open-sourced as a template design pattern. We opted to develop an Android application that detects plant diseases.

Challenges we ran into

Our greatest challenge has always been how to expand our platform and reach more smallholder farmers as most of them are in extremely remote locations and spread across the country. Scaling our initiatives will mean investments in expanding our server infrastructure, cloud computing platforms, trainers, and field officers.

Accomplishments that we're proud of

During the ideation phase, we were able as a team to come up with a solution that will overcome the number one problem in developing countries in Africa.

What we learned

What's next for FarmAi

On ensuring that the smallholder farmers are not locked out of the Farm AI App because of financial constraints, The Farm ai team has introduced a program that allows them to serve a small village (20 – 50 farmers) with one smartphone. Farmers without a smartphone can opt for their farms to be scanned by assigned ‘farmer leaders’. This ensures that a lot of farmers can access the Foundation's tools and services.

Built With

Share this project:

Updates