Bachelor Project
Machine Learning & Data Generation
A procedural environment generation framework in Unity3D with two convolutional neural network algorithms in Python.
Machine Learning is so very dependent on working with as much data as possible, and in the area of field robotics it is no different. This project had its focus on weed detection and classification and how it could be improved by providing an algorithm synthetically generated data, in the form of images, that correlated to a real world setting. The framework that was developed in Unity was used to generate these synthesised images, which were then exported and used to detect the so-called broad-leaved dock. To evaluate the image quality of the synthesised images, two different algorithms were applied, where one was executed in Matlab and the other executed in Google Colab. The results evaluating the image quality of the synthesised images showed to be promising, yielding to further investigation.
Keywords: AI & Machine Learning, Computer Grapics, Image Processing, 3D-modelling