BZPEERReading, trust and collaboration for research
Menu

Discover

GraphFollow connections around a paper, topic or saved view.

Workspaces

HomePick up where your research flow left off.InboxHandle requests, replies and notifications from one place.CollectionsCurate reading lists, evidence sets and shareable dossiers.ResearchSee your private provenance graph, trail and imports.CollaborationMove from discovery into focused discussion rooms.PublishDraft, preview and publish full-text research articles.

Network

ExpertsFind specialists worth following or asking for help.CommunitiesJoin research circles organized around shared work.PartnersSee external organizations active around the same topics.

Opportunities

ListingsBrowse roles, calls and collaborations with clearer fit signals.

Account

Log inReturn to your reading and collaboration workspace.Create accountStart saving work, following people and joining teams.
Log inCreate account

Researcher profile

Evgenia Rusak

Evgenia Rusak contributes to research discovery and scholarly infrastructure.

ResearcherAffiliation not importedOpen to collaborate
Computer VisionMachine Learningphysics.atom-phphysics.optics

Trust snapshot

Quick read

Trust 15 - UnverifiedVerification L1Unclaimed author
3works
0followers
4topics
4close collaborators

Actions

Decide how to stay connected

Follow researcher0

Identity and collaboration

How to connect with this researcher

Claiming links this public author record to a researcher profile and unlocks direct collaboration workflows.

Log in to claim

Direct collaboration

Open a focused conversation when the fit is right

Claim this author entity first to unlock direct invitations.

Research graph

See the researcher in context

Open full explorer

Inspect adjacent work, topics, institutions and collaborators without jumping out to a separate graph page.

Building this graph slice

BZPEER is loading the nearby papers, people, topics and institutions for this page.

Published work

3 published item(s)

preprint2020arXiv

A simple way to make neural networks robust against diverse image corruptions

The human visual system is remarkably robust against a wide range of naturally occurring variations and corruptions like rain or snow. In contrast, the performance of modern image recognition models strongly degrades when evaluated on previously unseen corruptions. Here, we demonstrate that a simple but properly tuned training with additive Gaussian and Speckle noise generalizes surprisingly well to unseen corruptions, easily reaching the previous state of the art on the corruption benchmark ImageNet-C (with ResNet50) and on MNIST-C. We build on top of these strong baseline results and show that an adversarial training of the recognition model against uncorrelated worst-case noise distributions leads to an additional increase in performance. This regularization can be combined with previously proposed defense methods for further improvement.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Benchmarking Robustness in Object Detection: Autonomous Driving when Winter is Coming

The ability to detect objects regardless of image distortions or weather conditions is crucial for real-world applications of deep learning like autonomous driving. We here provide an easy-to-use benchmark to assess how object detection models perform when image quality degrades. The three resulting benchmark datasets, termed Pascal-C, Coco-C and Cityscapes-C, contain a large variety of image corruptions. We show that a range of standard object detection models suffer a severe performance loss on corrupted images (down to 30--60\% of the original performance). However, a simple data augmentation trick---stylizing the training images---leads to a substantial increase in robustness across corruption type, severity and dataset. We envision our comprehensive benchmark to track future progress towards building robust object detection models. Benchmark, code and data are publicly available.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Enhancement of and interference among higher order multipole transitions in molecules near a plasmonic nanoantenna

Spontaneous emission of quantum emitters can be modified by engineering their optical environment. This allows a resonant nanoantenna to significantly modify the radiative properties of a quantum emitter. In this article, we go beyond the common electric dipole approximation for the molecular electronic transition and take light-matter coupling through higher order multipoles into account. We investigate, by means of theory and numerical simulations, a strong enhancement of the magnetic dipole and electric quadrupole emission channels of a molecule adjacent to a plasmonic patch nanoantenna. While this on its own had been considered, the assumption in prior work usually has been that each molecular transition is dominated only by one of those multipolar emission channels. This leads naturally to the notion of discussing the modified emission in terms of a modified local density of states defined for each specific multipolar transition. In reality, this restricts the applicability of the approach, since specific molecular transitions occur via multiple multipolar pathways that have to be considered all at once. Here, we introduce a framework to study interference effects between higher order transitions in molecules by (a) a rigorous quantum-chemical calculation of their multipolar moments and (b) by a consecutive investigation of the transition rate upon coupling to an arbitrarily shaped nanoantenna. Based on that formalism we predict interference effects between these transition channels. This allows for a strong suppression of radiation by exploiting destructive interference. Our work suggests that placing a suitably chosen molecule at a well defined position and at a well defined orientation relative to a nanoantenna can fully suppress the transition probability.

0 citations0 reviewsNo code linkedOpen access