Monocular Depth Estimation of micro scale Images
Keywords
Loading...
Authors
Issue Date
2020-01-31
Language
en
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Depth estimation based on a single image is a hot topic in current computer vision
and deep learning research. Current work almost exclusively uses data depicting street
or indoor scenes to train and benchmark their methods. However, there are plenty of
other domains where depth information can be useful. I will elaborate how current
state of the art methods can be applied to the domain of semiconductor devices where
depth information can be useful to assess whether a device is damaged or faulty. I will
showcase the entire process starting with a pipeline to obtain ground truth data and
ending with bench marking several models. More precisely, I will train and evaluate
a supervised, a self supervised and a semi supervised neural network. Ultimately,
I demonstrate that it is possible to extract some depth information but that more
research and data is needed to achieve dense and accurate disparity maps.
Description
Citation
Supervisor
Faculty
Faculteit der Sociale Wetenschappen