MwogoNet is an Artificial Intelligence project that uses Deep Learning. It aims at helping Cassava farmers and researchers diagnose diseases based on the signs portrayed on the leaf.
Cassava is the most widely grown food crop in Uganda, and Africa at large, and is source of carbohydrates. However, Cassava has one major weakness. It's susceptible to various cassava diseases. We are working on an automated method, using current state of the art Deep learning techniques to make it easier for farmers and researchers to detect and diagnose these diseases faster.
We think that to make our cassava disease diagnosis models more accurate, we need to first guide it, to learn what characteristics make up a leaf. Using semantic segmentation, MwogoNet first automatically extracts the noisy background from the leaf, and passes it to the Convolutional Neural Network trained to classify Cassava Mosaic Disease, Cassava Brownstreak disease and Cassava Bacterial Blight
Step1: Feed the model with unseen diseased (test)image
Step2: Model eliminates noisy background from the image through semantic segmentation
Step3: Model passes the new clear image to the disease classification Neural network
Step4: Model is able to classify the cassava diseases, returning predicted disease + Severity score
MwogoNet can be used by anyone curious and ready to use technology to ease the tedious and time consuming task of cassava disease diagnois. Farmers, Agricultural researchers, Students, Data Scientist and everyone is welcome to try out MwogoNet
Using Artificial intelligence in Africa, for Africa.
