Depth Map Reconstruction Using Deep Convolutional Neural Networks

Student: Korinevskaya Alisa

Supervisor: Ilya Makarov

Educational Programme: Data Science (Master)

Year of Graduation: 2018

In this paper, the problem of depth reconstruction is considered in terms of single depth map interpolation as well as super-resolution. In particular, the complex interpolating method based on the application of an interpolating convolutional network followed by a super-resolving one is proposed. It is shown that this technique is able to produce fast approximate depth maps. Furthermore, the super-resolving convolutional network trained with the Perceptual and GAN losses is suggested. It appears that the results obtained with this network are comparable with the state-of-the-art approaches in terms of SSIM, RMSE and PSNR metrics. The implemented methods are tested and evaluated on the SYNTHIA, NYU Depth V2, Matterport, KITTI, Sintel and Middlebury datasets.

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