direkt zum Inhalt springen

direkt zum Hauptnavigationsmenü

Sie sind hier

TU Berlin

Inhalt des Dokuments

Dr.-Ing. Tobias Senst

Research Assistant

Technische Universität Berlin
Telecommunication Systems
Communication Systems Group
Office E-N 1
Einsteinufer 17
10587 Berlin

Raum: E-N 340

Tel.: +49 30 314 - 28256
Fax: +49 30 314 - 22514

E-Mail: 
www: Linkedin ResearchGate  GoogleScholar StackOverflow GitHub

Research Activities

Robust Local Optical Flow Estimation

Bild

The Robust Local Optical Flow (RLOF) is a sparse optical flow and feature tracking method. The main objective is to provide a fast and accurate motion estimation solution. The main advantage of the RLOF approach is the adjustable runtime and computational complexity which is in contrast to most common optical flow methods linearly dependend on the number of motion vectors (features) to be estimated. Thus the RLOF is a local optical flow method and most related to the PLK method ( better known as KLT Tracker ) and thus the famous Lucas Kanade method. The sparse-to-dense interpolation scheme allows for fast computation of dense optical flow fields. mehr zu: Robust Local Optical Flow Estimation

Lagrangian-based Video Analytics

Bild

We aim for innovative ways to process and use dynamic patterns in video motion to quantify salient motion features and thus improve computer vision performance for tasks such as identification, segmentation, and classification. The proposed methodology provides a powerful set of data-driven descriptors for continuous and integral motion analysis on variable temporal scales (i.e., for short-term as well as long-term motion features). mehr zu: Lagrangian-based Video Analytics

Hyper-Parameter Optimization for Convolutional Neural Networks Committees based on Evolutionary Algorithms

Bild

We propose an evolutionary algorithm-based framework to automatically optimize the CNN structure by means of hyper-parameters. Further, we extend our framework towards a joint optimization of a committee of CNNs to leverage specialization and cooperation among the individual networks. Experimental results show a significant improvement over the state-of-the-art on the well-established MNIST dataset for hand-written digits recognition. mehr zu: Hyper-Parameter Optimization for Convolutional Neural Networks Committees based on Evolutionary Algorithms

People Carrying Object Detection and Classification

Bild

Detecting people carrying objects detection and classification is a problem known from surveillance scenarios. It can be used as a first step in order to monitor interactions between people and objects, like depositing or removing an object. Research is focused on new machine learning approaches for pedestrian detection and new ways of feature representation, behavior analysis and machine learning techniques for classification. mehr zu: People Carrying Object Detection and Classification

Privacy Preserving Technologies

Bild

The ever increasing number of surveillance camera networks being deployed all over the world has resulted in a high interest in the development of algorithms to automatically analyze the video footage, but has also opened new questions as how to efficiently manage the vast amount of information generated and, more important, how to protect the privacy of the individuals being recorded in their daily life. An technical approach are the Privacy Preserving Technologies (PET) which are algorithm to filter personal privacy related information and/or adapt the protection level of personal privacy in videos. mehr zu: Privacy Preserving Technologies

IOU Tracker

Bild

Tracking-by-detection is a common approach to multi-object tracking. With ever increasing performances of object detectors, the basis for a tracker becomes much more reliable. In combination with commonly higher frame rates, this poses a shift in the challenges for a successful tracker. We propose a very simple tracking algorithm which can compete with more sophisticated approaches at a fraction of the computational cost. With thorough experiments we show its potential using a wide range of object detectors. The proposed method can easily run at thousands of frames per second (fps) while outperforming the state-of-the-art on the DETRAC vehicle tracking dataset and achieves competitive results on the MOT17 benchmark. mehr zu: IOU Tracker

Background Substraction / Foreground Detection / Static-Object Detection

Bild

Gaussian mixture models have been extensively used and enhanced in the surveillance domain because of their ability to adaptively describe multimodal distributions in real-time with low memory requirements. Nevertheless, they still often suffer from the problem of converging to poor solutions if the main mode stretches and thus over-dominates weaker distributions. We propose complementary background models for background modelling and to detect static and moving objects in crowded video sequences. mehr zu: Background Substraction / Foreground Detection / Static-Object Detection

Awards

Best Paper Award @ IET ICDP 2015

Bild

We are delighted to announce that our paper "A Local Feature based on Lagrangian Measures for Violent Video Classification" won the Best Paper Award at the IET International Conference on Imaging for Crime Detection and Prevention, 15.07.2015 - 17.07.2015. Congratulations to Tobias Senst and the co-authors. mehr zu: Best Paper Award @ IET ICDP 2015

We won the VisDrone 2018 Challenge @ ECCV!

Bild

We are delighted to announce that our V-IOU tracker won the VisDrone 2018 Challenge for multi-object tracking at the the ECCV 2018 workshop "Vision Meets Drone: A Challenge" (or VisDrone2018, for short) on September 8, 2018, in Munich, Germany. mehr zu: We won the VisDrone 2018 Challenge @ ECCV!

Challenge Winner IWOT4S @ AVSS 2018

Bild

We are delighted to announce that our IOU tracker won in a row the IWOT4S Challenge for multi-object tracking at the International Workshop on Traffic and Street Surveillance for Safety and Security at IEEE AVSS 2018, Auchkland, New Zealand, 27.11.2018 mehr zu: Challenge Winner IWOT4S @ AVSS 2018

Ongoing Research Projects

SiGroViD

Bild

The aim of this project is to support security personnel at big events with remotely controlled drone-cameras monitoring the entire location, to automatically 3D reconstruct analyze the captured data and effectively visualize and highlight special events like violent behavior, emergence of panic, etc. In relationship to traditional installed cameras, drones allow a more flexible and comprehensive overview of the terrain, provide better analysis results and additionally reduce costs compared to the operation of helicopters. mehr zu: SiGroViD

Completed Research Projects

QuaVideo

Bild

QuaVideo - Video quality improvement based on long-term temporal trajectory filtering. Funded by BMBF (01.10.2013 - 30.11.2016). With the convergence of internet and digital television, low quality internet video is now frequently displayed on ultra high resolution, high quality displays. Typical examples include low resolution and poor quality Youtube video and older amateur family video displayed on 4K devices. Typical MPEG coding artifacts, like blocking and smearing, appear very annoying when displayed at high resolutions. Our challenge is the quality enhancement of legacy video to make content more appealing when viewed at large size displays. To this end we track motion information through hundreds of frames in the video scene to filter out noise and smearing artifacts. Our approach adapts to scene changes and varying content fully automatically and removes artifacts from hours of video. mehr zu: QuaVideo

LASIE

Bild

EU FP7 funded project LASIE - Large scale information exploitation of forensic data (01.05.2014 - 31.10.2017). The LASIE project aims to design and implement an open and expandable framework that will significantly increase the efficiency of current investigation practices, by providing an automated initial analysis of the vast amounts of heterogeneous forensic data that analysts have to cope with. The framework will be able to handle forensic data that have been acquired from a variety of different sources including CCTV surveillance content, confiscated desktops and hard disks, mobile devices, Internet, social networks, handwritten and calligraphic documents. mehr zu: LASIE

Mosaic

Bild

The goal of the MOSAIC project has been to develop the MOSAIC Platform. The MOSAIC Platform involves multi-modal data intelligence capture and analytics including video and text collaterals etc. The distributed intelligence within the platform enables decision support for automated detection, recognition, geo-location and mapping, including intelligent decision support at various levels to enhance situation awareness, surveillance targeting and camera handover; these involve level one fusion, and situation understanding to enable decision support and impact analysis at level two and three of situation assessment. mehr zu: Mosaic

VideoSense

Bild

EU FP7 funded project VideoSense (2011-2015). VideoSense will integrate leading European research groups to create a long-term open integration of critical mass in the twin areas of Ethically-Guided, and, Privacy Preserving Video Analytics where the advent of new data intelligence technologies against the background of dynamic societal and citizen’s goals, norms, expectations, safety and security needs have all contributed to a complex interplay of influences which deserve in-depth study and solution seeking in order for the European society, citizen and industry to strike the optimal balance in resolution of the various challenges in this arena. mehr zu: VideoSense

SinoVE

Bild

Projektpartner: •Bundespolizei, Potsdam •Deutsche Bahn AG, Berlin •Fraunhofer-Institut für Nachrichtentechnik, Heinrich-Hertz-Institut, Berlin •Funkwerk plettac electronic GmbH, Fürth •Gesellschaft zur Förderung angewandter Informatik, Berlin •Siemens Division Building Technologies, Karlsruhe •Technische Universität Berlin, Fachgebiet Nachrichtenübertragung, Berlin •VCS Video Communication Systems AG, BOSCH Group, Ottobrunn •Vis-à-pix GmbH, Potsdam Weitere Informationen: Projektbeschreibung Präsentation Abschlussberichte der Teilvorhaben mehr zu: SinoVE

Publications (Journals Papers)

2017

2012

Nach oben

Publications (Conference Papers)

2019

2018

  • Erik Bochinski, Tobias Senst, Thomas Sikora
    Extending IOU Based Multi-Object Tracking by Visual Information
    IEEE International Conference on Advanced Video and Signals-based Surveillance, 27.11.2018 - 30.11.2018, pp. 441-446
    ISBN: 978-1-5386-9294-3/18
    Details BibTeX
  • Gregory Schröder, Tobias Senst, Erik Bochinski, Thomas Sikora
    Optical Flow Dataset and Benchmark for Visual Crowd Analysis
    IEEE International Conference on Advanced Video and Signals-based Surveillance, Auckland, New Zealand, 27.11.2018 - 30.11.2018, pp. 7-11
    ISBN: 978-1-5386-9294-3/18
    Details BibTeX
  • Siwei Lyu, Ming-Ching Chang, Dawei Du, Wenbo Li, Yi Wei, Marco Del Coco, Pierluigi Carcagnì, Arne Schumann, Bharti Munjal, Dinh-Quoc-Trung Dang, Doo-Hyun Choi, Erik Bochinski, Fabio Galasso, Filiz Bunyak, Guna Seetharaman,Jang-Woon Baek, Jong Taek Lee, Kannappan Palaniappan, Kil-Taek Lim, Kiyoung Moon, Kwang-Ju Kim, Lars Sommer, Meltem Brandlmaier, Min-Sung Kang, Moongu Jeon, Noor M. Al-Shakarji, Oliver Acatay, Pyong-Kun Kim, Sikandar Amin, Thomas Sikora, Tien Dinh, Tobias Senst, Vu-Gia-Hy Che, Young-Chul Lim, Young-min Song, and Yun-Su Chung
    UA-DETRAC 2018: Report of AVSS2018 & IWT4S Challenge on Advanced Traffic Monitoring
    IEEE International Conference on Advanced Video and Signals-based Surveillance, Auckland, New Zealand, 27.11.2018 - 30.11.2018
    ISBN: 978-1-5386-9294-3/18
    Details BibTeX
  • Pengfei Zhu, Longyin Wen, Dawei Du, Xiao Bian, Haibin Ling, Qinghua Hu, Hao Cheng, Chengfeng Liu, Xiaoyu Liu, Wenya Ma, Qinqin Nie, Haotian Wu, Lianjie Wang, Arne Schumann, Dan Wang, Diego Ortego, Elena Luna, Emmanouil Michail, Erik Bochinski, Feng Ni, Filiz Bunyak, Gege Zhang, Guna Seetharaman, Guorong Li, Hongyang Yu, Ioannis Kompatsiaris, Jianfei Zhao, Jie Gao, Jose Martinez, Juan Miguel, Kannappan Palaniappan, Konstantinos Avgerinakis, Lars Sommer, Martin Lauer, Mengkun Liu, Noor Al-Shakarji, Oliver Acatay, Panagiotis Giannakeris, Qijie Zhao, Qinghua Ma, Qingming Huang, Stefanos Vrochidis, Thomas Sikora, Tobias Senst, Wei Song, Wei Tian, Wenhua Zhang, Yanyun Zhao, Yidong Bai, Yinan Wu, Yongtao Wang, Yuxuan Li, Zhaoliang Pi, Zhiming Ma
    VisDrone-VDT2018: The Vision Meets Drone Video Detection and Tracking Challenge Results
    ECCV 2018 Workshops, volume Part V, Munich, Germany, 08.09.2018 - 14.09.2018, pp. 496-518
    Details BibTeX

2017

2016

2015

2014

2013

2012

  • Tobias Senst, Rubén Heras Evangelio, Ivo Keller, Thomas Sikora
    Clustering Motion for Real-Time Optical Flow based Tracking
    IEEE International Conference on Advanced Video and Signal-Based Surveillance (AVSS 2012), Beijing, China, 18.09.2012 - 21.09.2012, pp. 410--415
    ISBN: 978-1-4673-2499-1 DOI: 10.1109/AVSS.2012.20
    Details BibTeX
  • Tobias Senst, Alexander Kuhn, Holger Theisel, Thomas Sikora
    Detecting People Carrying Objects utilizing Lagrangian Dynamics
    IEEE International Conference on Advanced Video and Signal-Based Surveillance (AVSS 2012), Beijing, China, 18.09.2012 - 21.09.2012, pp. 398--403
    ISBN: 978-1-4673-2499-1 DOI: 10.1109/AVSS.2012.34
    Details BibTeX
  • Esra Acar, Tobias Senst, Alexander Kuhn, Ivo Keller, Holger Theisel, Sahin Albayrak, Thomas Sikora
    Human Action Recognition using Lagrangian Descriptors
    IEEE Workshop on Multimedia Signal Processing (MMSP 2012), Banff, Canada, 17.09.2012 - 19.09.2012, pp. 360--365
    IEEE Catalog Number: CFP12MSP-USB E-ISBN : 978-1-4673-4571-2 Print ISBN: 978-1-4673-4570-5 INSPEC Accession Number: 13116360 DOI:10.1109/MMSP.2012.6343469
    Details BibTeX
  • Alexander Kuhn, Tobias Senst, Ivo Keller, Thomas Sikora, Holger Theisel
    A Lagrangian Framework for Video Analytics
    IEEE Workshop on Multimedia Signal Processing (MMSP 2012), Banff, Canada, 17.09.2012 - 19.09.2012, pp. 387--392
    IEEE Catalog Number: CFP12MSP-USB E-ISBN : 978-1-4673-4571-2 Print ISBN: 978-1-4673-4570-5 INSPEC Accession Number: 13116365 DOI: 10.1109/MMSP.2012.6343474
    Details BibTeX
  • Tobias Senst, Brigitte Unger, Ivo Keller, Thomas Sikora
    Performance Evaluation of Feature Detection for Local Optical Flow Tracking
    International Conference on Pattern Recognition Applications and Methods (ICPRAM 2012), volume 2, Vilamoura, Portugal, 06.02.2012 - 08.02.2012, pp. 303--309
    DOI: 10.5220/0003731103030309
    Details BibTeX

2011

2010

2009

Nach oben

Software & Datasets

Robust Local Optical Flow Library (RLOF) @ OpenCV contrib (4.1)

Bild

The Robust Local Optical Flow (RLOF) is a sparse optical flow and feature tracking method. We are deligthed that it is now part of OpenCV Contribution library (4.1.0). The RLOF methods are motivated by the problem of local motion estimation via robust regression with linear models. The main objective is to provide real-time capability, accurate and scaleable motion estimation solution. The software implements several versions of the RLOF algorithms for sparse and dense optical flow estimation. mehr zu: Robust Local Optical Flow Library (RLOF) @ OpenCV contrib (4.1)

TUB CrowdFlow Dataset

Bild

Optical Flow Dataset and Benchmark for Visual Crowd Analysis. A new optical flow dataset exploiting the possibilities of a recent video engine to generate sequences with groundtruth optical flow for large crowds in different scenarios. We break with the development of the last decade of introducing ever increasing displacements to pose new difficulties. Instead we focus on real-world surveillance scenarios where numerous small, partly independent, non rigidly moving objects observed over a long temporal range pose a challenge. mehr zu: TUB CrowdFlow Dataset

Bachelor- und Masterthesis

2019

  • Kenan Karalioglu
    Deep Learning Ansatz für die Klassifikation von Videodaten mittels Lagrang`schen Maßen
    01.08.2019, bachelor thesis tutored by Dr.-Ing. Tobias Senst, Prof. Dr.-Ing. Thomas Sikora
    Details BibTeX
  • Yang Xu
    StereoBrush: Combining Interactive Depth Map Creation with Optical Flow Estimation
    19.02.2019, bachelor thesis tutored by Dr.-Ing. Sebastian Knorr/Dipl.-Ing. Tobias Senst, Prof. Dr.-Ing. Thomas Sikora
    Details BibTeX
  • Yanbo Zhu
    Evaluation of Feature and Segmentation-Based Registration Techniques to Estimate Object Dimensions from Drone Videos
    30.01.2019, master thesis tutored by Dipl.-Ing. Tobias Senst, Prof. Dr.-Ing. Thomas Sikora
    Details BibTeX

2018

  • Alexander Metzler
    Tracking von Einzelpersonen in dichten Menschenmengen
    21.11.2018, bachelor thesis tutored by Erik Bochinski, M.Sc./Dipl.-Ing. Tobias Senst, Prof. Dr.-Ing. Thomas Sikora
    Details BibTeX
  • Leandro Fagin
    Untersuchung von Methoden zur Täuschung von CNN-basierten Gesichtsdetektionsverfahren
    14.11.2018, bachelor thesis tutored by Dipl.-Ing. Tobias Senst/ Erik Bochinski, M.Sc., Prof. Dr.-Ing. Thomas Sikora
    Details BibTeX
  • Gregory Peter Schröder
    A New Optical Flow Dataset and Benchmark for Visual Crowd Analysis
    07.11.2018, master thesis tutored by Dipl.-Ing. Tobias Senst, Prof. Dr.-Ing. Thomas Sikora
    Details BibTeX

2017

  • Robert Schröder
    Partielle globale Bewegungsmodelle für lokale Optische Fluss Verfahren
    01.11.2017, bachelor thesis tutored by Dipl.-Ing. Tobias Senst, Prof. Dr.-Ing. Thomas Sikora
    Details BibTeX

2016

  • Tim Barnewski
    Eine Analyse der Steering Kernel Regression in Abhängigkeit der Stützstellen für die Bildverarbeitung
    02.06.2016, master thesis tutored by Dipl.-Ing. Tobias Senst, Prof. Dr.-Ing. Thomas Sikora
    Details BibTeX
  • Jonas Geistert
    Interpolation von Bewegungsvektoren für die Erstellung von dichten optischen Flussfeldern
    21.01.2016, master thesis tutored by Dipl.-Ing. Tobias Senst, Prof. Dr.-Ing. Thomas Sikora
    Details BibTeX
  • Ghassen Bacha
    Untersuchung von kernelbasierten Kodebücher für die Detektion von Gewalt in Videos
    07.01.2016, bachelor thesis tutored by Dipl.-Ing. Tobias Senst, Prof. Dr.-Ing. Thomas Sikora
    Details BibTeX

2015

  • Dimitri Bilostotskyy
    Skalen-abhängige Optimierung der Abbruchkriterien des Lucas/Kanade Verfahrens
    19.11.2015, bachelor thesis tutored by Dipl.-Ing. Tobias Senst, Prof. Dr.-Ing. Thomas Sikora
    Details BibTeX
  • Peter Ullmann
    Trajektorienberechnung von Merkmalspunkten in Videosequenzen anhand von lokalen und globalen Bewegungsmodellen
    10.11.2015, bachelor thesis tutored by Dipl.-Ing. Tobias Senst/Dipl.-Ing. volker Eiselein, Prof. Dr.-Ing. Thomas Sikora
    Details BibTeX
  • Patrick Florian Krusch
    Generation of High Definition Video Material using Super Resolution
    25.06.2015, bachelor thesis tutored by Dipl.-Ing. Tobias Senst, Prof. Dr.-Ing. Thomas Sikora
    Details BibTeX

2014

  • Brigitte Unger
    Detektion von Nanopartikeln in REM-Bildsequenzen von Makrophagen
    16.01.2014, master thesis tutored by Dipl.-Ing. Tobias Senst, Prof. Dr.-Ing. Thomas Sikora
    Details BibTeX

2013

  • Florian Lackorn
    Realisierung eines Semi-Automatischen Label-Werkzeugs für Videoobjekte mithilfe von Trackingverfahren
    30.04.2013, bachelor thesis tutored by Dipl.-Ing. Tobias Senst, Dr.-Ing. Kristian Weiß, Carmeq GmbH, Prof. Dr.-Ing. Thomas Sikora
    Details BibTeX
  • Sabrina Blümling
    Feature Tracking using the Robust Local Optical Flow
    14.02.2013, master thesis tutored by Dipl.-Ing. Tobias Senst, Prof. Dr.-Ing. Thomas Sikora
    Details BibTeX

2012

  • Jonas Geistert
    Untersuchung des Konvergenzverhaltens des pyramidalen Lucas/Kanade Verfahrens
    28.11.2012, bachelor thesis tutored by Dipl.-Ing. Tobias Senst, Prof. Dr.-Ing. Thomas Sikora
    Details BibTeX
  • Jens Krenzin
    Analyse von Personenströmen mit statistikbasierten Vektorfeldern
    16.07.2012, bachelor thesis tutored by Dipl.-Ing. Tobias Senst, Dipl.-Ing. Volker Eiselein, Prof. Dr.-Ing. Thomas Sikora
    Details BibTeX

Zusatzinformationen / Extras