Mobile application with a neural network Corteva Expert

Scanner for recognition of weeds, pests and diseases of agricultural crops

Task

Train a neural network on an existing dataset assembled by Corteva employees. Create an application for the company's customers through which they will be able to recognize bugs, diseases and weeds in rapeseed, sunflower and corn fields.

Solution

We analyzed lots of existing solutions for neural networks, stopped on Turicreate (based on ResNet), and made the given dataset. Then we trained three neural networks: for bugs, weeds and pests. We developed instructions for users to create photos for recognition inside the application in order to increase the accuracy of the result.