Search Results for author: Ivan Sosnovik

Found 13 papers, 8 papers with code

Learning to Summarize Videos by Contrasting Clips

no code implementations12 Jan 2023 Ivan Sosnovik, Artem Moskalev, Cees Kaandorp, Arnold Smeulders

Video summarization aims at choosing parts of a video that narrate a story as close as possible to the original one.

Contrastive Learning Unsupervised Video Summarization

LieGG: Studying Learned Lie Group Generators

1 code implementation9 Oct 2022 Artem Moskalev, Anna Sepliarskaia, Ivan Sosnovik, Arnold Smeulders

Symmetries built into a neural network have appeared to be very beneficial for a wide range of tasks as it saves the data to learn them.

Contrasting quadratic assignments for set-based representation learning

1 code implementation31 May 2022 Artem Moskalev, Ivan Sosnovik, Volker Fischer, Arnold Smeulders

The views are ordered in pairs, such that they are either positive, encoding different views of the same object, or negative, corresponding to views of different objects.

Contrastive Learning Metric Learning +1

Wiggling Weights to Improve the Robustness of Classifiers

1 code implementation18 Nov 2021 Sadaf Gulshad, Ivan Sosnovik, Arnold Smeulders

To demonstrate that wiggling the weights consistently improves classification, we choose a standard network and modify it to a transform-augmented network.

PIE: Pseudo-Invertible Encoder

no code implementations ICLR 2019 Jan Jetze Beitler, Ivan Sosnovik, Arnold Smeulders

We consider the problem of information compression from high dimensional data.

Two is a crowd: tracking relations in videos

no code implementations11 Aug 2021 Artem Moskalev, Ivan Sosnovik, Arnold Smeulders

Tracking multiple objects individually differs from tracking groups of related objects.

Object Relation +1

How to Transform Kernels for Scale-Convolutions

no code implementations ICCVW 2021 Ivan Sosnovik, Artem Moskalev, Arnold Smeulders

We aim for accurate scale-equivariant convolutional neural networks (SE-CNNs) applicable for problems where high granularity of scale and small kernel sizes are required.

Built-in Elastic Transformations for Improved Robustness

no code implementations20 Jul 2021 Sadaf Gulshad, Ivan Sosnovik, Arnold Smeulders

We focus on building robustness in the convolutions of neural visual classifiers, especially against natural perturbations like elastic deformations, occlusions and Gaussian noise.

Data Augmentation

DISCO: accurate Discrete Scale Convolutions

1 code implementation4 Jun 2021 Ivan Sosnovik, Artem Moskalev, Arnold Smeulders

In recent work scale equivariance was added to convolutional neural networks.

Scale Equivariance Improves Siamese Tracking

1 code implementation17 Jul 2020 Ivan Sosnovik, Artem Moskalev, Arnold Smeulders

We develop the theory for scale-equivariant Siamese trackers, and provide a simple recipe for how to make a wide range of existing trackers scale-equivariant.

Translation Visual Object Tracking +1

Scale-Equivariant Steerable Networks

1 code implementation ICLR 2020 Ivan Sosnovik, Michał Szmaja, Arnold Smeulders

The effectiveness of Convolutional Neural Networks (CNNs) has been substantially attributed to their built-in property of translation equivariance.

Computational Efficiency Image Classification +1

Neural networks for topology optimization

2 code implementations27 Sep 2017 Ivan Sosnovik, Ivan Oseledets

The main novelty of this work is to state the problem as an image segmentation task.

Decoder Image Segmentation +1

Cannot find the paper you are looking for? You can Submit a new open access paper.