Hello, I'm Diana Sungatullina

I am a third year PhD student at the Applied Algebra & Geometry Group at Czech Technical University in Prague, supervised by Prof. Tomas Pajdla jointly with Prof. Konrad Schindler via the ELLIS PhD Program. I received my Specialist degree (eq. to BSc.+MSc.) in Applied Mathematics and Computer Science from Lomonosov Moscow State University, Moscow, Russia in 2014. My research interests lie in Computer Vision, Multiple View Geometry, and Deep Learning.

CV Scholar Github

Publications

MinBackProp — Backpropagating through Minimal Solvers

Diana Sungatullina, Tomas Pajdla

Journal of WSCG, 2024

An efficient and robust approach to backpropagate through minimal problem solvers in end-to-end training.

PDF Code

Image Manipulation with Perceptual Discriminators

Diana Sungatullina*, Egor Zakharov*, Dmitry Ulyanov, Victor Lempitsky

ECCV, 2018

A new architecture that combines perceptual and adversarial losses in non-additive manner for image translation.

Project Page PDF Code

Photorealistic Monocular Gaze Redirection Using Machine Learning

Daniil Kononenko, Yaroslav Ganin, Diana Sungatullina, Victor Lempitsky

TPAMI, 2018

Three approaches to the gaze redirection problem in images using machine learning.

PDF

DeepWarp: Photorealistic Image Resynthesis for Gaze Manipulation

Yaroslav Ganin, Daniil Kononenko, Diana Sungatullina, Victor Lempitsky

ECCV, 2016

A new deep architecture that performs coarse-to-fine warping with intensity correction, enabling gaze redirection across a range of angles.

Project Page PDF Code

Fast registration algorithms for histological images

Diana Sungatullina, Andrey Krylov, Dmitry Fedorov

Scientific Visualization, 2014

Fast contour-based matching with minimal loss of quality for histological images.

Project Page

Multiview discriminative learning for age-invariant face recognition

Diana Sungatullina, Jiwen Lu, Gang Wang, Pierre Moulin

FG, 2013

A discriminative learning method with multiview feature representations that projects different types of local features into a latent subspace in a specialized manner.

PDF