Project: Visual distinction of plants by machine learning
Manual weeding of root vegetables is a crucial task for frugal crop yields. However, it is an expensive, lengthy and difficult task for farmers. Desherbex aims at building a robotic farm equipment that can replace manual weeding. It implies that the tool must 1) distinguish weeds from root vegetables and 2) eliminate weeds with robotic tools. This internship will be working mostly on the first point. The project is to design a neural network with machine learning that will be able to identify plants on a video feed. The output of this neural network will be used to display a preliminary result on a graphical user interface to help farmers and field workers identify quickly the weeds to be eliminated by the farm equipment. This visual identification must be done quickly and on an embedded system, to control the robotic weeding tools. The video feed, the output of the neural network and the interaction with the operator will be recorded to build better data models to improve the algorithm that distinguishes weeds and root vegetables.
Supervisor(s):
Partner(s)
