Visual positioning in indoor environments

Project Leader: Kourosh Khoshelham
Primary Contact: Kourosh Khoshelham (k.khoshelham@unimelb.edu.au)
Keywords: computer vision; deep learning; geomatics; machine learning; spatial information
Disciplines: Infrastructure Engineering
Domains:

In the absence of GPS signals in indoor environments, visual positioning using imagery captured by smartphones offers a promising solution for indoor positioning and navigation. The state of the art visual positioning methods are based on machine learning using a dataset of images with known pose. Such images are generally not available in indoor environments but can be generated synthetically from a 3D model of the environment. The aim of this research is to investigate generative methods (eg, CycleGAN) for the generation of photorealistic synthetic images from a 3D model.