• A
  • A
  • A
  • ABC
  • ABC
  • ABC
  • А
  • А
  • А
  • А
  • А
Regular version of the site

Modeling Associativity with Neural Networks

ФИО студента: Ilya Solovev

Руководитель: Kirill O. Neklyudov

Кампус/факультет: Faculty of Computer Science

Программа: Data Science (Master)

Год защиты: 2020

Solving modern computer vision and deep learning problems often requires large labeled datasets. Meanwhile, open-source data could be used for feature extraction and transfer learning. In this work we propose a new method of training based on the visual context of objects. Our method uses data with only labeled bounding boxes and masks and could be generalized for unlabeled data. We then show that features extracted with our model are semantically meaningful and could be successfully used for further training a method for a classification task. Moreover, our method has a natural ability to map appropriate surroundings with an object and vice-versa. We also present an approach based on our model for matching a word with related images without any explicit labels. That can be used to simplify manual labeling or to get suitable images by a natural language request. Thus, we model "associativity", providing an ability to predict most likely scenes and objects or words and images. Keywords: feature extraction, latent representation, deep learning, computer vision, word2vec.

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