Disparity Estimation via Neural Networks

Student: Bogomolov Pavel

Educational Programme: Applied Mathematics and Information Science (Bachelor)

Final Grade: 7

Year of Graduation: 2019

Recent works in supervised depth estimation rely on datasets with the ground truth depth measured via sensors alongside the images. However, the quality of such measurements is usually low due to various hardware limitations; they are also expen- sive to obtain. We can mitigate these shortcomings by switching to a semi-supervised setting. Several methods have been proposed which utilize stereo cameras or video sequences as supervision. Our approach uses the stereo setup and is based heavily on the Monodepth framework by Godard et al. with a substantial change: we re- place the network architecture with Double Refinement Network. This allows us to achieve much better results in depth estimation.

Full text (added May 19, 2019)

Student Theses at HSE must be completed in accordance with the University Rules and regulations specified by each educational programme.

Summaries of all theses must be published and made freely available on the HSE website.

The full text of a thesis can be published in open access on the HSE website only if the authoring student (copyright holder) agrees, or, if the thesis was written by a team of students, if all the co-authors (copyright holders) agree. After a thesis is published on the HSE website, it obtains the status of an online publication.

Student theses are objects of copyright and their use is subject to limitations in accordance with the Russian Federation’s law on intellectual property.

In the event that a thesis is quoted or otherwise used, reference to the author’s name and the source of quotation is required.

Search all student theses