Eleonora Vig

Eleonóra Víg

Senior Applied Scientist


I am a senior computer vision scientist at Amazon, in Berlin, Germany. My research interests include saliency prediction, multi-object detection and tracking, and human action recognition, using deep learning techniques and the simulation of virtual worlds.

LinkedIn profile
Google Scholar profile

About me

I received my PhD in Computer Vision from the University of Lübeck, Germany, for work on human gaze prediction and guidance in videos. In the following two years, I was a postdoctoral research fellow in the Computer and Biological Vision Lab of David Cox at the Center for Brain Science, Harvard University. During my Post-Doc, I was partly funded by a German Academic Exchange Service (DAAD) grant. From 2013 to 2016, I worked as a research scientist in the Computer Vision group at Xerox Research Centre Europe (now Naver Labs Europe), in Grenoble, France. Between 2016 and 2019, I was with the Remote Sensing Technology Institute of the German Aerospace Center (DLR) working on aerial computer vision, including human crowd analysis and the modeling of traffic participants for autonomous driving. Since Sept. 2019, I am a senior applied scientist at Amazon.


Selected publications

My Google Scholar profile.

Y. Hou, E. Vig, M. Donoser, L. Bazzani. Learning Attribute-driven Disentangled Representations for Interactive Fashion Retrieval. ICCV, 2021.
[paper] [code]

S. M. Azimi, C. Henry, L. Sommer, A. Schumann, E. Vig. SkyScapes - Fine-Grained Semantic Understanding of Aerial Scenes. ICCV, 2019.

S. M. Azimi, E. Vig, R. Bahmanyar, M. Körner, P. Reinartz. Towards Multi-class Object Detection in Unconstrained Remote Sensing Imagery. ACCV, 2018.

C. de Souza, A. Gaidon, E. Vig, A. M. Lopez. Sympathy for the Details: Dense Trajectories and Hybrid Classification Architectures for Action Recognition. ECCV, 2016.

S. Jetley, N. Murray, E. Vig. End-To-End Saliency Mapping via Probability Distribution Prediction. CVPR, 2016. (spotlight presentation)
[pdf] [bibtex]

A. Gaidon, Q. Wang, Y. Cabon, E. Vig. Virtual Worlds as Proxy for Multi-Object Tracking Analysis. CVPR, 2016.
[pdf] [bibtex] [dataset] [poster]

A. Gaidon, E. Vig. Online Domain Adaptation for Multi-Object Tracking. BMVC, 2015. (oral)
[pdf] [bibtex]

E. Vig, M. Dorr, D. Cox. Large-Scale Optimization of Hierarchical Features for Saliency Prediction in Natural Images. CVPR, 2014.
[pdf] [bibtex] [code] [poster]

M. Milford, W. Scheirer, E. Vig, A. Glover, O. Baumann, J. Mattingley, D. Cox. Condition-Invariant, Top-Down Visual Place Recognition. ICRA, 2014.
[pdf] [bibtex]

M. Milford, E. Vig, W. Scheirer, D. Cox. Vision‐based Simultaneous Localization and Mapping in Changing Outdoor Environments. Journal of Field Robotics 31 (5), 780-802, 2014.
[pdf] [bibtex]

E. Vig, M. Dorr, D. Cox. Space-variant Descriptor Sampling for Action Recognition based on Saliency and Eye Movements. ECCV, 2012.
[pdf] [bibtex] [dataset] [poster]

E. Vig, M. Dorr, T. Martinetz, E. Barth. Intrinsic Dimensionality Predicts the Saliency of Natural Dynamic Scenes. IEEE Transactions on Pattern Analysis and Machine Intelligence (T-PAMI) 34 (6), 1080-1091, 2012.
[pdf] [bibtex]

E. Vig, M. Dorr, E. Barth. Efficient Visual Coding and the Predictability of Eye Movements on Natural Movies. Spatial Vision 22 (5), 397-408, 2009.
[pdf] [bibtex]

Software and datasets

Virtual KITTI
A photo-realistic synthetic video dataset designed to learn and evaluate computer vision models for several video understanding tasks, such as object detection and multi-object tracking, scene-level and instance-level semantic segmentation, optical flow, and depth estimation. Link to the accompanying CVPR'2016 paper.
eDN saliency
Reference code for computing Ensembles of Deep Networks (eDN) saliency maps based on the CVPR'2014 paper "Large-Scale Optimization of Hierarchical Features for Saliency Prediction in Natural Images".
Eye movement dataset for the Hollywood2 benchmark
A large dataset of eye movements we collected from five subjects who performed the action recognition task for the Hollywood2 Action Recognition challenge. The accompanying ECCV'2012 paper can be found here.

Theme adapted from orderedlist/minimal.