Selected Publications

If you are interested in one of the following publications and no pdf- or ps-file is available or you have problems downloading it, please send email to Bernt Schiele

2008

2007

Towards Robust Pedestrian Detection in Crowded Image Sequences
Edgar Seemann, Mario Fritz, Bernt Schiele
in IEEE Conference on Computer Vision and Pattern Recognition (CVPR'07), Minneapolis, USA, 2007

Cross-Modal Learning of Visual Categories using Different Levels of Supervision
Mario Fritz, Geert-Jan Kruijff, Bernt Schiele
in International Conference on Vision Systems (ICVS'07), Bielefeld, Germany, 2007

Multi-class Classification with Dependent Gaussian Processes
Mykhaylo Andriluka, Lorenz Weizsaecker, Thomas Hofmann
in 12th International Conference on Applied Stochastic Models and Data Analysis, Crete, Greece, 2007.

2006

2005

Integrating Representative and Discriminant Models for Object Category Detection. Mario Fritz, Bastian Leibe, Barbara Caputo, and Bernt Schiele. In International Conference on Computer Vision (ICCV'05), Beijing, China, October 2005.

Local Features for Object Class Recognition. Krystian Mikolajczyk, Bastian Leibe, and Bernt Schiele. In International Conference on Computer Vision (ICCV'05), Beijing, China, October 2005.

An Evaluation of Local Shape-Based Features for Pedestrian Detection. Edgar Seemann, Bastian Leibe, Krystian Mikolajczyk, and Bernt Schiele. Accepted for oral presentation in British Machine Vision Conference (BMVC'05), Oxford, UK, September 2005.

Context Annotation for A Live Life Recording. Nicky Kern, Albrecht Schmidt, Bernt Schiele. In Personal and Ubiquitous Computing

Towards improving trust in context-aware systems by displaying system confidence. Stavros Antifakos, Nicky Kern, Adrian Schwaninger, Bernt Schiele. In MobileHCI, Salzburg, Austria, 2005.

Pedestrian Detection in Crowded Scenes. Bastian Leibe, Edgar Seemann, and Bernt Schiele. Accepted for oral presentation in IEEE International Conference on Computer Vision and Pattern Recognition (CVPR'05), San Diego, CA, June 2005.

Sensing and Monitoring Professional Skiing Athletes: Lessons Learned from a Collaboration with Ski Trainer's. Florian Michahelles and Bernt Schiele, In Pervasive Computing Magazin, IEEE, July-September, Special Issue on Pervasive Computing in Sports, 2005.

Experiencing Technology before it exists: A Case Study. Florian Michahelles and Bernt Schiele. In What makes for good application-led research in ubiquitous computing?, A Pervasive 2005 Workshop, Munich, Germany, May 2005.

Analyzing Features for Activity Recognition. Tâm Huynh and Bernt Schiele. In Proceedings of the 2005 joint conference on Smart objects and ambient intelligence (EUSAI), October 2005, Grenoble, France, ACM Press New York, NY, USA. (bibtex)

2004

Towards Distributed Awareness - An Artifact based Approach. Florian Michahelles,Stavros Antifakos, Albrecht Schmidt, Michael Beigl, and Bernt Schiele. In Sixth IEEE Workshop on Mobile Computing Systems and Applications (WMCSA 2004), Lake District, UK, Dec 2004.

A Model for Human Interruptability: Experimental Evaluation and Automatic Estimation from Wearable Sensors, Nicky Kern, Stavros Antifakos, Bernt Schiele, Adrian Schwaninger In 8th International Symposium on Wearable Computing (ISWC), Washington DC, USA, November 2004.

Less Contact: Heart-rate detection without even touching the user. Florian Michahelles, Ramon Wicki and Bernt Schiele. In Eighth IEEE International Symposium on Wearable Computers Conference (ISWC), Arlington, USA 2004.

Evaluating the Effects of Displaying Uncertainty in Context-Aware Applications. Stavros Antifakos, Adrian Schwaninger, and Bernt Schiele. In Ubicomp'04. 6th International Conference on Ubiquitous Computing. Nottingham, UK, Sept. 2004.

A Semantic Typicality Measure for Natural Scene Categorization Julia Vogel and Bernt Schiele. In DAGM'04 Annual Pattern Recognition Symposium Tuebingen, Germany, Aug. 2004.

Scale Invariant Object Categorization Using a Scale-Adaptive Mean-Shift Search. Bastian Leibe and Bernt Schiele. In DAGM'04 Annual Pattern Recognition Symposium, Tuebingen, Germany, Aug. 2004. Springer LNCS, Vol. 3175, pp. 145-153, 2004. DAGM Best Paper Award.

Natural Scene Retrieval based on a Semantic Modeling Step Julia Vogel and Bernt Schiele. In Conference on Image and Video Retrieval CIVR 2004, Dublin, Ireland, July 2004.

Combined Object Categorization and Segmentation with an Implicit Shape Model. Bastian Leibe, Ales Leonardis, and Bernt Schiele. In ECCV'04 Workshop on Statistical Learning in Computer Vision, Prague, May 2004.

Towards Situation-Aware Affordances: An Experimental Study. Stavros Antifakos, Florian Michahelles, and Bernt Schiele. In Pervasive'04 International Conference on Pervasive Computing, Vienna, April 2004.

Building intelligent environments with Smart-Its. Lars Erik Holmquist, Hans-Werner Gellersen, Gerd Kortuem, Albrecht Schmidt, Martin Strohbach, Stavros Antifakos, Florian Michahelles, Bernt Schiele , Michael Beigl , Ramia Maze. Computer Graphics and Applications, IEEE, Jan/Feb 2004, page 56 - 64, Volume: 24 , Issue: 1, ISSN: 0272-1716 .

2003

Fast and Robust Face Finding via Local Context. Hannes Kruppa, Modesto Castrillon-Santana and Bernt Schiele. Joint IEEE International Workshop on Visual Surveillance and Performance Evaluation of Tracking and Surveillance (VS-PETS'03), Nice, France, October 2003.

Automatic Detection and Tracking of Abandoned Objects. Martin Spengler and Bernt Schiele. Joint IEEE International Workshop on Visual Surveillance and Performance Evaluation of Tracking and Surveillance (VS-PETS'03), Nice, France, October 2003.

The Role of Perception in Ubiquitous Interactive Systems. Stavros Antifakos, Jan Borchers and Bernt Schiele, At the Crossroads: The Interaction of HCI and System Issues in Ubicomp. Workshop at: The Fifth International Conference on Ubiquitous Computing (Ubicomp), Seattle, USA, October 2003.

Instructions immersed into the real world How your Furniture can teach you. Florian Michahelles, Stavros Antifakos, Jani Boutellier, Albrecht Schmidt and Bernt Schiele, Poster Submission, The Fifth International Conference on Ubiquitous Computing (Ubicomp), Seattle, USA, October 2003.

LaughingLily: Using a Flower as a Real World Information Display. Stavros Antifakos and Bernt Schiele, Poster Submission, The Fifth International Conference on Ubiquitous Computing (Ubicomp), Seattle, USA, October 2003.

Grouping Mechanisms for Smart Objects Based on Implicit Interaction and Context Proximity. Stavros Antifakos and Bernt Schiele, Poster Submission, The Fifth International Conference on Ubiquitous Computing (Ubicomp), Seattle, USA, October 2003.

A-Life: Saving Lives in Avalanches. Florian Michahelles and Bernt Schiele, Video Submission, The Fifth International Conference on Ubiquitous Computing (Ubicomp), Seattle, USA, October 2003.

Designing Physical Interaction with Sensor Drawbacks in Mind. Stavros Antifakos, Jan Borchers, and Bernt Schiele, Physical Interaction 2003 Workshop on Real World User Interfaces (PI03), Fifth International Symposium on Human-Computer Interaction with Mobile Devices and Services (Mobile HCI 2003), Udine, Italy, September 8-11, 2003.

Sensing Opportunities for Physical Interaction. Florian Michahelles and Bernt Schiele, Physical Interaction 2003 Workshop on Real World User Interfaces (PI03), Fifth International Symposium on Human-Computer Interaction with Mobile Devices and Services (Mobile HCI 2003), Udine, Italy, September 8-11, 2003.

Applying Wearable Sensors to Avalanche Rescue: First Experiences with a Novel Avalanche Beacon. Florian Michahelles, Peter Matter, Albrecht Schmidt and Bernt Schiele, In Computers & Graphics, Vol. 27, No. 6, 2003.

Wearable Sensing to Annotate Meeting Recordings Nicky Kern, Bernt Schiele, Holger Junker, Paul Lukowicz, and Gerhard Tröster. In Personal and Ubiquitous Computing: Selected papers from the ISWC2002 Conference, 2003.

Using Local Context To Improve Face Detection Hannes Kruppa and Bernt Schiele, In British Machine Vision Conference (BMVC'03), Norwich, UK, Sept. 2003.

Interleaved Object Categorization and Segmentation. Bastian Leibe and Bernt Schiele, In British Machine Vision Conference (BMVC'03), Norwich, UK, Sept. 2003.

Designing Physical Interaction with Sensor Drawbacks in Mind. Stavros Antifakos, Jan Borchers, and Bernt Schiele, PI2003 Physical Interaction 2003 Workshop on Real World User Interfaces, Mobile HCI 2003 Fifth International Symposium on Human-Computer Interaction with Mobile Devices and Services (Udine, Italy, September 8�11, 2003).

Multi-Sensor Activity Context Detection for Wearable Computing. Nicky Kern, Bernt Schiele and Albrecht Schmidt, In European Symposium on Ambient Intelligence(EUSAI), Eindhoven, The Netherlands, 2003.

Towards an Inertial Sensor Network. Kristof van Laerhoven, Nicky Kern, Hans-Werner Gellersen, and Bernt Schiele, In IEE EuroWearable 2003 (EuroWearable), Birmingham, UK, 2003.

Context-Aware Notifications for Wearable Computing. Nicky Kern and Bernt Schiele, In International Symposium on Wearable Computing (ISWC), New York, USA, 2003.

Applying Wearable Sensors to Avalanche Rescue: First Experiences with a Novel Avalanche Beacon. Florian Michahelles, Peter Matter, Albrecht Schmidt and Bernt Schiele, In International Workshop on Mobile Computing "Assistance, Mobility, Applications" (IMC Workshop), Rostock, Germany, 2003.

Analyzing Appearance and Contour Based Methods for Object Categorization. Bastian Leibe and Bernt Schiele, In International Conference on Computer Vision and Pattern Recognition (CVPR'03), Madison, Wisconsin, June 2003.

A-Life - Increasing Survival Chances in Avalanches by Wearable Sensors. Florian Michahelles, Timo Ahonen and Bernt Schiele, In 3rd International Workshop on Smart Appliances and Wearable Computing (IWSAWC 2003), Providence, Rhode Island USA, 2003, ISSN 1432-7864.

Multi-Object Tracking Based on a Modular Knowledge Hierarchy. Martin Spengler and Bernt Schiele, In 3th International Conference on Computer Vision Systems ICVS'03, Graz, Austria, April 2003.

Towards Robust Multi-Cue Integration for Visual Tracking. Martin Spengler and Bernt Schiele, In Machine Vision and Applications, Vol. 14(1), pp. 50-58, Springer, 2003

The Perceptive Workbench: Computer-Vision based Gesture Tracking, Object Tracking, and 3D Reconstruction for Augmented Desks. Thad Starner, Bastian Leibe, David Minnen, Tracy Westyn, Amy Hurst, Justin Weeks, In Machine Vision and Applications, Vol. 14(1), pp. 59-71, Springer, 2003.

2002

Better Rescue through Sensors. Florian Michahelles, Bernt Schiele, In First International Workshop on Ubiquitous Computing for Cognitive Aids (UbiCog'02), at UbiComp 2002, G�teborg (Gothenburg), Sweden, 2002.

Wearable Sensing to Annotate Meeting Recordings Nicky Kern, Bernt Schiele, Holger Junker, Paul Lukowicz, and Gerhard Tröster. In The 6th International Symposium on Wearable Computers, ISWC 2002, Seattle, Washington, USA, October 2002.

Bridging the Gap Between Virtual and Physical Games using Wearable Sensors Stavros Antifakos and Bernt Schiele, To Appear: In The 6th International Symposium on Wearable Computers, ISWC 2002, Seattle, Washington, USA, October 2002.

Proactive Instructions for Furniture Assembly Stavros Antifakos, Florian Michahelles and Bernt Schiele, To Appear: In The Fourth International Conference on Ubiquitous Computing, UbiComp 2002, Göteborg, Sweden, September 2002.

Multi-Object Tracking: Explicit Knowledge Representation and Implementation for Complexity Reduction Martin Spengler and Bernt Schiele, In Cognitive Vision Workshop 2002, Zurich, Switzerland, September 2002.

Query-Dependent Performance Optimization for Vocabulary-Supported Image Retrieval Julia Vogel and Bernt Schiele, In DAGM Symposium 2002, Zurich, Switzerland, September 2002.

Skin Patch Detection in Real-World Images Hannes Kruppa, Martin A. Bauer, and Bernt Schiele, To Appear: In DAGM Symposium 2002, Zurich, Switzerland, September 2002.

Saliency of Interest Points under Scale Changes. Daniela Hall, Bastian Leibe, and Bernt Schiele, To Appear: In British Machine Vision Conference 2002, Cardiff, Wales, September 2002.

Smart CAPs for Smart Its - Context Detection for Mobile Users Florian Michahelles, Michael Samulowitz, Personal and Ubiquitous Computing, Volume 6, Issue 4 (2002), pp 269-275, Springer, London, September 2002.

How Many Classifiers Do I Need? Bernt Schiele. To Appear in International Conference on Pattern Recognition ICPR 2002, Quebec, Canada, August 2002.

Analyzing Non-Negative Matrix Factorization for Image Classification David Guillamet, Bernt Schiele, and Jordi Vitri. To Appear in International Conference on Pattern Recognition ICPR 2002, Quebec, Canada, August 2002.

On Performance Characterization and Optimization for Image Retrieval Julia Vogel and Bernt Schiele, In European Conference on Computer Vision ECCV 2002, Copenhagen, Denmark, May 2002.

Mutual Information For Evidence Fusion Hannes Kruppa and Bernt Schiele. In Section "Sensor Fusion, Registration and Planning" (http://www.dai.ed.ac.uk/CVonline/fusion.htm) of CVonline: On-Line Compendium of Computer Vision [Online]. R. Fisher (ed), January 2002.

Detecting Context in Distributed Sensor Networks by Using Smart Context-Aware Packets. Florian Michahelles, Michael Samulowitz and Bernt Schiele, In: International Conference on Architecture of Computing Systems (ARCS), 2002, Karlsruhe, Germany, April 2002.

2001

Computer Vision Systems, Bernt Schiele and Gerhard Sagerer (Eds.), Lecture Notes in Computer Science 2095, Springer, 2001.

3D Object Recognition from Range Images using Local Feature Histograms. Guenter Hetzel, Bastian Leibe, Paul Levi, and Bernt Schiele, In: International Conference on Computer Vision and Pattern Recognition CVPR 2001, Kauai Island, Hawaii, December 2001.

Beyond Position Awareness. Bernt Schiele and Stavros Antifakos, In: Proceedings of the Workshop on Location Modeling at Ubicomp 2001, October 2001.

Performance Prediction for Vocabulary-Supported Image Retrieval Julia Vogel and Bernt Schiele, In: International Conference on Image Processing ICIP 2001, Thessaloniki, Greece, October 2001.

Smart-Its Friends: A Technique for Users to Easily Establish Connections between Smart Artefacts Lars Erik Holmquist, Friedemann Mattern, Bernt Schiele, Petteri Alahuhta, Michael Beigl, Hans-W. Gellersen, To appear in: Ubicomp 2001, Atlanta, Gorgia, USA, October 2001.

Smart CAPs for Smart Its - Context Detection for Mobile Users. Florian Michahelles, Michael Samulowitz, In Third International Workshop on Human Computer Interaction with Mobile Devices (MobileHCI), at IHM-HCI 2001, Lille, France, September 2001.

Hierarchical Combination of Object Models using Mutual Information Hannes Kruppa and Bernt Schiele, To appear in: British Machine Vision Conference, BMVC 2001, Manchester, UK, September 2001.

Sensory Augmented Computing and its Potential for Human-Computer Interaction. Bernt Schiele, To appear in: HCI International 2001, New Orleans, USA, August 2001.

Context-driven Model Switching For Visual Tracking Hannes Kruppa, Martin Spengler, and Bernt Schiele. In 9th International Symposium on Intelligent Robotic Systems 2001, Toulouse, France, July 2001.

Towards Robust Multi-cue Integrtion for Visual Tracking MartinSpengler and Bernt Schiele. In 9th International Workshop of Computer Vision Systems, 2001, Vancouver, Canda, July 2001.

Sensory Augmented Computing: Wearing the Museum's Guide Bernt Schiele, T. Jebara, and N. Oliver. In: IEEE Micro May/June 2001.

Towards Robust Perception and Model Integration Bernt Schiele, Martin Spengler, and Hannes Kruppa. To appear in: Sensor Based Intelligent Robots, Lecture Notes in Computer Science, Springer, 2001.

2000

Vocabulary-Supported Image Retrieval Bernt Schiele and Julia Vogel, In 1st DELOS Workshop on Information Seeking, Searching and Querying in Digital Libraries, Zurich, Switzerland, December 2000.

Using Mutual Information to Combine Object Models Hannes Kruppa and Bernt Schiele. In 8th International Symposium on Intelligent Robotic Systems 2000, Reading, UK, July 2000.

Towards Automatic Extraction and Modelling of Objects from Image Sequences Bernt Schiele. In 8th International Symposium on Intelligent Robotic Systems 2000, Reading, UK, July 2000.

Recognition without Correspondence using Multidimensional Receptive Field Histograms. Bernt Schiele and James L. Crowley. IJCV - International Journal of Computer Vision. 36(1), p 31-50, January 2000

1999

Probabilistic Object Recognition and Localization. Bernt Schiele and Alex Pentland. In ICCV'99 International Conference on Computer Vision.

Learning Audio-Visual Associations Using Mutual Information Deb Roy, Bernt Schiele and Alex Pentland. In Integration of Speech and Image Understanding, Sept 1999

An Interactive Computer Vision System - DyPERS: Dynamic Personal Enhanced Reality System. Bernt Schiele, Nuria Oliver, Tony Jebara and Alex Pentland. In ICVS'99 International Conference on Vision Systems.

Situation Aware Computing with Wearable Computers. Bernt Schiele, Thad Starner, Brad Rhodes, Brian Clarkson and Alex Pentland. Chapter in forthcoming book Augmented Reality and Wearable Computers , W. Barfield and T. Caudell (editors), Lawrence Erlbaum Press.

1998

Visual Context Awareness via Wearable Computing. Thad Starner, Bernt Schiele and Alex Pentland. In ISWC'98 International Symposium on Wearable Computers.

Recognizing Places using Image Sequences. Hisashi Aoki, Bernt Schiele and Alex Pentland. In PUI'98 Perceptual User Interfaces.

Augmented Realities Integrating User and Physical Model. Thad Starner, Bernt Schiele, Bradley Rhodes, Tony Jebara, Nuri Oliver, Josh Weaver and Alex Pentland. In IWAR'98 International Workshop on Augmented Reality.

Comprehensive Colour Image Normalization Graham D. Finlayson, Bernt Schiele and James L. Crowley. In ECCV'98 Fifth European Conference on Computer Vision.

Using colour for image indexing Graham D. Finlayson, Bernt Schiele and James L. Crowley. In The Challenge of Image Retrieval.

Transinformation for Active Object Recognition Bernt Schiele and James L. Crowley. In ICCV'98, International Conference on Computer Vision, Bombay, India, January 1998

1997

Recognition without Correspondence using Multidimensional Receptive Field Histograms Bernt Schiele and James L. Crowley. In IJCV - International Journal of Computer Vision. 36(1), p 31-50, January 2000 . Also M.I.T. Media Laboratory, Perceptual Computing Section Technical Report No. 453, Dec 1997.

Object Recognition using Mutidimensional Receptive Field Histograms Bernt Schiele. PhD thesis, Institut National Polytechnique de Grenoble, english translation of the french thesis, July 1997.

Position Estimation for a Mobile Robot From Principal Components of Laser Range Data Frank Wallner, Bernt Schiele and James L. Crowley. In 5th International Symposium on Intelligent Robotic Systems'97, Stockholm, Sweden, July 1997. Also to appear in Robotics and Autonomous Systems, 1998.

Transinformation of Object Recognition and its Application to Viewpoint Planning Bernt Schiele and James L. Crowley. In Robotics and Autonomous Systems, Vol 21, No 1, July 1997.

The concept of Visual Classes for Object Classification. Bernt Schiele and James L. Crowley. In SCIA'97, Scandinavian Conference on Image Analysis, Lappeenranta, Finland, June 1997.

1996

Where to look next and what to look for. Bernt Schiele and James L. Crowley. In IROS'96, Intelligent Robots and Systems, Osaka, Japan, December 1996.

Probabilistic Object Recognition Using Multidimensional Receptive Field Histograms. Bernt Schiele and James L. Crowley. In ICPR'96, International Conference on Pattern Recognition, Vienna, Austria, August 1996.

Object Recognition Using Multidimensional Receptive Field Histograms. Bernt Schiele and James L. Crowley. In ECCV'96, Fourth European Conference on Computer Vision, Cambridge, UK, April 1996.

The Robustness of Object Recognition to View Point Changes Using Multidimensional Receptive Field Histograms Bernt Schiele and James L. Crowley. Presented at ECIS-VAP meeting, Object Recognition Day. Israel, March 1996.

1995

The Robustness of Object Recognition to Rotation Using Multidimensional Receptive Field Histograms Bernt Schiele and James L. Crowley. Presented at Rosenon Worshop on Computational Vision, July 1995.

Estimation of the Head Orientation based on a Face-Color-Intensifier Bernt Schiele and Alex Waibel. In 3rd International Symposium on Intelligent Robotic Systems'95, Pisa, Italy, July 1995.

Gaze Tracking Based on Face-Color Bernt Schiele and Alex Waibel. In International Workshop on Automatic Face- and Gesture-Recognition, , Zurich, Switzerland, June 1995.

1994

A Comparison of Position Estimation Techniques Using Occupancy Grids Bernt Schiele and James L. Crowley. In IEEE International Conference on Robotics and Automation, May 1994.

A Comparison of Position Estimation Techniques Using Occupancy Grids Bernt Schiele and James L. Crowley. In Robotics and Autonomous Systems, 1994

1993

Certainty Grids: Perception and Localization for a Mobile Robot Bernt Schiele and James L. Crowley. In International Workshop on Intelligent Robotics Systems'93, Zakopane, Poland, July 1993.

Paper abstracts and links to ps-files

An Evaluation of Local Shape-Based Features for Pedestrian Detection. Edgar Seemann, Bastian Leibe, Krystian Mikolajczyk, and Bernt Schiele. In British Machine Vision Conference (BMVC'05) Oxford, UK, September 2005. Accepted for oral presentation. pdf.

Abstract: Pedestrian detection in real world scenes is a challenging problem. In recent years a variety of approaches have been proposed, and impressive re- sults have been reported on a variety of databases. This paper systematically evaluates (1) various local shape descriptors, namely Shape Context and Lo- cal Chamfer descriptor and (2) four different interest point detectors for the detection of pedestrians. Those results are compared to the standard global Chamfer matching approach. A main result of the paper is that Shape Con- text trained on real edge images rather than on clean pedestrian silhouettes combined with the Hessian-Laplace detector outperforms all other tested ap- proaches.

Context Annotation for A Live Life Recording. Nicky Kern, Albrecht Schmidt, Bernt Schiele. In Personal and Ubiquitous Computing pdf.

Abstract: In the near future it will be possible to continuously record and store the entire audio-visual lifetime of a person together with all digital information that person perceives or creates. While the storage of this data will be possible soon, retrieval and indexing into such large data sets is an unsolved challenge. Since today.s retrieval cues seem insufficient we argue that additional cues, obtained from bodyworn sensors, make associative retrieval by humans possible. We present three approaches to create such cues, each along with an experimental evaluation: the users physical activity from acceleration sensors, his social environment from audio, and his interruptability from multiple sensors.

Towards improving trust in context-aware systems by displaying system confidence. Stavros Antifakos, Nicky Kern, Adrian Schwaninger, Bernt Schiele. In MobileHCI, Salzburg, Austria, 2005. pdf.

Abstract: For automatic or context-aware systems a major issue is user trust, which is to a large extent determined by system reliability. For systems based on sensor input which are inherently uncertain or even uncomplete there is little hope that they will ever be perfectly reliable. In this paper we test the hypothesis if explicitly displaying the current confidence of the system increases the usability of such systems. For the example of a context-aware mobile phone, the experiments show that displaying confidence information increases the user's trust in the system.

Pedestrian Detection in Crowded Scenes. Bastian Leibe, Edgar Seemann, and Bernt Schiele. In IEEE International Conference on Computer Vision and Pattern Recognition (CVPR'05), San Diego, CA, June 2005. Accepted for oral presentation. pdf.

Abstract: In this paper, we address the problem of detecting pedestrians in crowded real-world scenes with severe overlaps. Our basic premise is that this problem is too difficult for any type of model or feature alone. Instead, we present a novel algorithm that integrates evidence in multiple iterations and from different sources. The core part of our method is the combination of local and global cues via a probabilistic top-down segmentation. Altogether, this approach allows to examine and compare object hypotheses with high precision down to the pixel level. Qualitative and quantitative results on a large data set confirm that our method is able to reliably detect pedestrians in crowded scenes, even when they overlap and partially occlude each other. In addition, the flexible nature of our approach allows it to operate on very small training sets.

A Model for Human Interruptability: Experimental Evaluation and Automatic Estimation from Wearable Sensors Nicky Kern, Stavros Antifakos, Bernt Schiele, Adrian Schwaninger In 8th International Symposium on Wearable Computing (ISWC), Washington DC, USA, November 2004. pdf

Abstract: For the estimation of user interruptability in wearable and mobile settings, we propose in in [8] to distinguish between the users' personal and social interruptability. In this paper, we verify this thesis with a user study on 24 subjects. Results show that there is a significant difference between social and personal interruptability. Further, we present a novel approach to estimate the social and personal interruptability of a user from wearable sensors. It is scalable for a large number of sensors, contexts, and situations and allows for online adaptation during run-time. We have developed a wearable platform, that allows to record and process the data from a microphone, 12 body-worn 3D acceleration sensors, and a location estimation. We have evaluated the approach on three different data sets, with a maximal length of two days.

Scale Invariant Object Categorization Using a Scale-Adaptive Mean-Shift Search. Bastian Leibe and Bernt Schiele. In DAGM'04 Annual Pattern Recognition Symposium, Tuebingen, Germany, Aug. 2004. Springer LNCS, Vol. 3175, pp. 145-153, 2004. DAGM Best Paper Award. pdf.

Abstract: The goal of our work is object categorization in real-world scenes. That is, given a novel image we want to recognize and localize unseen-before objects based on their similarity to a learned object category. For use in a real-world system, it is important that this includes the ability to recognize objects at multiple scales.
In this paper, we present an approach to multi-scale object categorization using scale-invariant interest points and a scale-adaptive Mean-Shift search. The approach builds on the method from [Leibe & Schiele, 2003], which has been demonstrated to achieve excellent results for the single-scale case, and extends it to multiple scales. We present an experimental comparison of the influence of different interest point operators and quantitatively show the method's robustness to large scale changes.

Combined Object Categorization and Segmentation with an Implicit Shape Model. Bastian Leibe, Ales Leonardis, and Bernt Schiele. In ECCV'04 Workshop on Statistical Learning in Computer Vision, Prague, May 2004. pdf.

Abstract: We present a method for object categorization in real-world scenes. Following a common consensus in the field, we do not assume that a figure-ground segmentation is available prior to recognition. However, in contrast to most standard approaches for object class recognition, our approach effectively segments the object as a result of the categorization.
This combination of recognition and segmentation into one process is made possible by our use of an Implicit Shape Model, which integrates both into a common probabilistic framework. In addition to the recognition and segmentation result, it also generates a per- pixel confidence measure specifying the area that supports a hypothesis and how much it can be trusted. We use this confidence to derive a natural extension of the approach to handle multiple objects in a scene and resolve ambiguities between overlapping hypotheses with an MDL-based criterion.
In addition, we present an extensive evaluation of our method on a standard dataset for car detection and compare its performance to existing methods from the literature. Our results show that the proposed method significantly outperforms previously published methods while needing one order of magnitude less training examples. Finally, we present results for articulated objects, which show that the proposed method can categorize and segment unfamiliar objects in different articulations and with widely varying texture patterns, even under significant partial occlusion.

Fast and Robust Face Finding via Local Context. Hannes Kruppa, Modesto Castrillon-Santana and Bernt Schiele. Joint IEEE International Workshop on Visual Surveillance and Performance Evaluation of Tracking and Surveillance (VS-PETS'03), Nice, France, October 2003. pdf,

Abstract: In visual surveillance face detection can be an important cue for initializing tracking algorithms. Recent work in psychophics hints at the importance of the local context of a face for robust detection, such as head contours and torso. This paper describes a detector that actively utilizes the idea of local context. The promise is to gain robustness that goes beyond the capabilities of traditional face detection making it particularly interesting for surveillance. The performance of the proposed detector in terms of accuracy and speed is evaluated on data sets from PETS 2000 and PETS 2003 and compared to the object-centered approach. Particular attention is paid to the role of available image resolution.

Wearable Sensing to Annotate Meeting Recordings. , Nicky Kern, Bernt Schiele, Holger Junker, Paul Lukowicz, and Gerhard Tröster In Personal and Ubiquitous Computing: Selected papers from the ISWC2002 Conference, 2003. pdf, project webpage

Abstract: We propose to use wearable computers and sensor systems to generate personal contextual annotations in audio visual recordings of meetings. In this paper we argue that such annotations are essential and effective to allow retrieval of relevant information from large audio-visual databases. The paper proposes several useful annotations that can be derived from cheap and unobtrusive sensors. It also describes a hardware platform designed to implement this concept, outlines approaches to extract annotations and presents first experimental results.

Using Local Context to Improve Face Detection. Hannes Kruppa and Bernt Schiele, In British Machine Vision Conference (BMVC'03), Norwich, UK, Sept. 2003. pdf

Abstract: Most face detection algorithms locate faces by classifying the content of a detection window iterating over all positions and scales of the input image. Recent developments have accelerated this process up to real-time performance at high levels of accuracy. However, even the best of today's computational systems are far from being able to compete with the detection capabilities of the human visual system. Psychophysical experiments have shown the importance of local context in the face detection process. In this paper we investigate the role of local context for face detection algorithms. In experiments on two large data sets we find that using local context can significantly increase the number of correct detections, particularly in low resolution cases, uncommon poses or individual appearances as well as occlusions.

Interleaved Object Categorization and Segmentation. Bastian Leibe and Bernt Schiele, In British Machine Vision Conference (BMVC'03), Norwich, UK, Sept. 2003. pdf, project webpage

Abstract: Historically, figure-ground segmentation has been seen as an important and even necessary pre-cursor for object recognition. In that context, segmentation is mostly defined as a data driven, that is bottom-up, process. As for humans object recognition and segmentation are heavily intertwined processes, it has been argued that top-down knowledge from object recognition can and should be used for guiding the segmentation process. In this paper, we present a method for the categorization of unfamiliar objects in difficult real-world scenes. The method generates object hypotheses without prior segmentation, which can be used to obtain a category-specific figure-ground segmentation. In particular, the proposed approach uses a probabilistic formulation to incorporate knowledge about the recognized category as well as the supporting information in the image to segment the object from the background. This segmentation can then be used for hypothesis verification, to further improve recognition performance. Experimental results show the capacity of the approach to categorize and segment object categories as diverse as cars and cows.

Designing Physical Interaction with Sensor Drawbacks in Mind. Stavros Antifakos, Jan Borcher and Bernt Schiele, In PI03, Workshop on Physical Interaction at Mobile HCI 2003. Udine Italy. pdf

Abstract: Physical interaction often relies on information stemming from sensors perceiving the real world. Sensors however, have imperfections resulting in drawbacks such as uncertainty and latency. Consequently, the improvement of sensors and perception methods is important. In this paper we argue however that imperfections of sensing will remain and that the key to better physical interaction lies in taking into account those sensor drawbacks explicitly during the design of the interaction. In order to take a first step in this direction we analyze sensor drawbacks and their effects on physical interaction. Based on this discussion we propose example solutions to the arising problems.

Multi-Sensor Activity Context Detection for Wearable Computing. Nicky Kern, Bernt Schiele and Albrecht Schmidt, In European Symposium on Ambient Intelligence(EUSAI), Eindhoven, The Netherlands, 2003. pdf

Abstract: For wearable computing applications, human activity is a central part of the user�s context. In order to avoid user annoyance it should be acquired automatically using body-worn sensors. We propose to use multiple acceleration sensors that are distributed over the body, because they are lightweight, small and cheap. Furthermore activity can best be measured where it occurs. We present a hardware platform that we developed for the investigation of this issue and results as to where to place the sensors and how to extract the context information.

Towards an Inertial Sensor Network. Kristof van Laerhoven, Nicky Kern, Hans-Werner Gellersen, and Bernt Schiele, In IEE EuroWearable 2003 (EuroWearable), Birmingham, UK, 2003. pdf

Abstract: Wearable inertial sensors have become an inexpensive option to measure the movements and positions of a person. Other techniques that use environmental sensors such as ultrasound trackers or vision-based methods need full line of sight or a local setup, and it is complicated to access this data from a wearable computer�s perspective. However, a body-centric approach where sensor data is acquired and processed locally, has a need for appropriate algorithms that have to operate under restricted resources. The objective of this paper is to give an overview of algorithms that abstract inertial data from body-worn sensors, with verification using state-of-the-art wearable multi-accelerometer prototypes.

Context-Aware Notification for Wearable Computing. Nicky Kern and Bernt Schiele, In International Symposium on Wearable Computing (ISWC), New York, USA, 2003. pdf

Abstract: In this paper we propose to use context information obtained from body--worn sensors to mediate notifications for a wearable computer. In particular we introduce a model which uses two axes, namely personal and social interruptability of the user in order to decide both whether or not to notify the user and to decide which notification modality to use. Rather than to model and recognize the complete context of the user we argue that personal and social interruptability can be derived directly from various sensors by the combination of tendencies. First experimental results show the feasibility of the approach using acceleration, audio, and location sensors.

Applying Wearable Sensors to Avalanche Rescue: First Experiences with a Novel Avalanche Beacon. Florian Michahelles, Peter Matter, Albrecht Schmidt and Bernt Schiele, In International Workshop on Mobile Computing "Assistance, Mobility, Applications" (IMC Workshop), Rostock, Germany, 2003. pdf

Abstract: We present a novel approach to enhance avalanche companion rescue using wearable sensing technologies. The time to find and extricate victims is most crucial: once buried by an avalanche, survival chances drop dramatically already after the first 15 minutes. Current technology offers only information on the location of a single victim, however statistics show that in many case there are multiple victims. In our research we address this issue and also investigate how the use of wearable sensors can further enhance such devices. We report on experiments using sensors to measure vital signs and environmental conditions and their suitability for avalanche rescue. Visualization for this type of application is addressed and two design sketches, both visualizing multiple victims and urgency, are presented. The architecture of current devices is extended by components to integrate these additional functions. We developt a prototypical implementation of an avalanche beacon supporting multiple victims and visualization of vital signs. This prototype was used for further experiments and offered a basis for participatory evaluation with practitioners in the field. A short overview of these results is presented.

Analyzing Appearance and Contour Based Methods for Object Categorization. Bastian Leibe and Bernt Schiele, In International Conference on Computer Vision and Pattern Recognition (CVPR'03), Madison, Wisconsin, June 2003. pdf, project webpage

Abstract: Object recognition has reached a level where we can identify a large number of previously seen and known objects. However, the more challenging and important task of categorizing previously unseen objects remains largely unsolved. Traditionally, contour and shape based methods are regarded most adequate for handling the generalization requirements needed for this task. Appearance based methods, on the other hand, have been successful in object identification and detection scenarios. Today little work is done to systematically compare existing methods and characterize their relative capabilities for categorizing objects.
In order to compare different methods we present a new database specifically tailored to the task of object categorization. It contains high-resolution color images of 80 objects from 8 different categories, for a total of 3280 images. It is used to analyze the performance of several appearance and contour based methods. Interestingly, the best categorization result is obtained by an appropriate combination of different methods.

A-Life - Increasing Survival Chances in Avalanches by Wearable Sensors.Florian Michahelles, Timo Ahonen, Bernt Schiele, In 3rd International Workshop on Smart Appliances and Wearable Computing (IWSAWC 2003), at ICDCS 2003, Providence, Rhode Island USA, 2003, ISSN 1432-7864. pdf

Abstract: Avalanches are one of the major threats to life in high mountain terrain. Once buried by an avalanche, survival chances dramatically drop from 92% after 15 minutes to only 30% after 35 minutes mostly due to the lack of oxygen. It is therefore extremely important to rescue any victims as fast as possible in order to maximize survival chances. Today's technology, so called electronic avalanche beacons, only allow to localize buried victims. In this paper we propose a novel avalanche rescue system enhanced with wearable sensors. Those sensors provide information about the vital state of buried victims such as heart rate, respiration activity, and blood oxygen saturation, as well as the orientation of the victim. We believe that this knowledge can empower non-professional companion rescuers with a tool to perform triage, i.e. sorting victims into categories of priority for treatment. Better allocation of resources can help to maximize survival chances of avalanche victims.

Multi-Object Tracking Based on a Modular Knowledge Hierarchy. Martin Spengler and Bernt Schiele, In 3th International Conference on Computer Vision Systems ICVS'03, Graz, Austria, April 2003. pdf, SpringerLink

Abstract: An important goal of research in computer vision systems is to develop architectures which are general and robust and at the same time transparent and easily transferable from one domain to another. To this extent this paper discusses and demonstrates the versatility of a multi-object tracking framework based on the so called knowledge hierarchy. The systematic description and analysis of a priori knowledge provides means not only for reducing the complexity of the multi-object tracking problem but also for building modular systems for solving it. The modularity of the framework, an essential ingredient for versatility, allows to replace individual parts of an existing system without altering the rest of the system or the overall architecture. The paper presents the modular framework including the knowledge hierarchy for multi object tracking. In order to demonstrate the transferability of the proposed approach the tracking framework is then applied to three different tracking scenarios (parking lot surveillance, people interaction monitoring, and dining table setup).

Towards Robust Multi-Cue Integration for Visual Tracking. Martin Spengler and Bernt Schiele, In Machine Vision and Applications, Vol 14(1), pp. 50-58, Springer, 2003. pdf, SpringerLink

Abstract: Even though many of today's vision algorithms are very successful,they lack robustness since they are typically tailored to a particular situation. In this paper we argue that the principles of sensor and model integration can increase the robustness of today's computer vision systems substantially. As an example, multi-cue tracking of faces is discussed. The approach is based on the principles of self-organization of the integration mechanism and self-adaptation of the cue models during tracking. Experiments show that the robustness of simple models is leveraged significantly by sensor and model integration.

The Perceptive Workbench: Computer-Vision Based Gesture Tracking, Object Tracking, and 3D Reconstruction for Augmented Desks. Thad Starner, Bastian Leibe, David Minnen, Tracy Westyn, Amy Hurst, and Justin Weeks, In Machine Vision and Applications, Vol 14(1), pp. 59-71, Springer, 2003. pdf, SpringerLink

Abstract: The Perceptive Workbench endeavors to create a spontaneous and unimpeded interface between the physical and virtual worlds.Its vision-based methods for interaction constitute an alternative to wired input devices and tethered tracking. Objects are recognized and tracked when placed on the display surface. By using multiple infrared light sources, the object's 3D shape can be
captured and inserted into the virtual interface. This ability permits spontaneity since either preloaded objects or those objects
selected at run-time by the user can become physical icons. Integrated into the same vision-based interface is the ability to
identify 3D hand position, pointing direction, and sweeping arm gestures. Such gestures can enhance selection, manipulation, and
navigation tasks. The Perceptive Workbench has been used for a variety of applications, including augmented reality gaming and
terrain navigation. This paper focuses on the techniques used in implementing the Perceptive Workbench and the system's performance.

Better Rescue through Sensors. Florian Michahelles, Bernt Schiele, In First International Workshop on Ubiquitous Computing for Cognitive Aids (UbiCog'02), at UbiCom 2002, G�teborg (Gothenburg), Sweden, 2002. pdf

Abstract: This paper investigates how wearable sensing technology can be applied to alpine avalanche rescue. Sensors worn by mountaineers reveal information about vital functions of the victims in an emergency case. In particular the rescue of multiple victims might benefit enormously from the use of wearable sensors. As time is the most critical issue in rescue, we believe that based on those sensor information rescuers obtain the necessary information in order to focus on victims which are still alive rather than to waste valuable time to dig out victims which are already dead.

Wearable Sensing to Annotate Meeting Recordings, Nicky Kern, Bernt Schiele, Holger Junker, Paul Lukowicz, and Gerhard Tröster, In The 6th International Symposium on Wearable Computers, Seattle, Washington, USA, October 2002. pdf

Abstract: We propose to use wearable computers and sensor systems to generate personal contextual annotations in audio visual recordings of meetings. Such annotations are essential to faciliate efficient retrieval of relevant information from large audio visual databases. The paper describes how useful annotations can be derived from cheap and unobtrusive sensors. It describes a hardware platform designed to implement this concept and presents first experimental results.

Bridging the Gap Between Virtual and Physical Games using Wearable Sensors, Stavros Antifakos and Bernt Schiele, To Appear: In The 6th International Symposium on Wearable Computers, Seattle, Washington, USA, October 2002. pdf

Abstract: We believe that game playing in the real world finds such large appreciation because it is based on interaction between people in a physical environment. In this paper we present an example of how the gap between virtual and physical games can be bridged using sensing technology from a wearable computer. Unmasking Mister X is a game we propose, which incorporates sensor data from all the players. It is a first step in enhancing real world games with wearable computers and sensing technology.

Proactive Instructions for Furniture Assembly, Stavros Antifakos, Florian Michahelles and Bernt Schiele, To Appear: In The Fourth International Conference on Ubiquitous Computing, UbiComp 2002, Göteborg, Sweden, September 2002. short, extended,Video

Abstract: Tennenhouse [1] coined the term proactive computing where humans get out of the interaction loop and may be serviced speci cally according to their needs and current situation. In this paper we propose a framework for proactive guidance which aims to overcome limitations of today's printed instructions. By attaching computing devices and multi- ple sensors onto di erent parts of the assembly the system can recognize the actions of the user and determine the current state of the assembly. The system can suggest the next most appropriate action at any point in time. In an experimental case study with the IKEA PAX wardrobe we show the feasibility of the proposed approach. At the end important issues are discussed and future directions are outlined.

Multi-Object Tracking: Explicit Knowledge Representation and Implementation for Complexity Reduction, Martin Spengler and Bernt Schiele. In Cognitive Vision Workshop 2002, Zurich, Switzerland, September 2002. pdf

Abstract: Many successful single-object tracking algorithms are formulated or may be reformulated as Bayesian inference problem. It is straight-forward to generalize the Bayesian formulation to the problem of multi-object tracking. However, due to the increase in dimensionality this formulation also opens Pandora's box in terms of exponential explosion of the computational complexity. In this paper we propose to constraint the computational complexity by exploiting and explicitly using prior knowledge at various levels of the Bayesian formulation of multi-object tracking. More specifically we discuss the use of a knowledge hierarchy which makes explicit where and how to introduce available knowledge.

Query-Dependent Performance Optimization for Vocabulary-Supported Image Retrieval, Julia Vogel and Bernt Schiele. In DAGM Symposium 2002, Zurich, Switzerland, September 2002. pdf

Abstract: Performance characterization of content-based image retrieval (CBIR) systems is especially difficult, because performance depends not only on the users, but also the tasks and the applications. In this paper, we propose a query-dependent performance characterization and optimization. The user specifies a high-level concept to be searched for, the size of the image region to be covered by the concept and an optimization constraint. Possible constraints might be 'maximum recall, 'maximum precision' or 'joint maximization of precision and recall'. The optimization proceeds in two stages. In the first stage, the detector best satisfying the user query is selected of a multitude of concept detectors. In the second stage, the information of the detectors is combined and optimized in order to reach optimum performance. Besides the optimization procedure itself the paper discusses the generation of multiple classifiers. In experiments, the advantage of jointly optimizing the query interval and the concept detector selection is shown.

Skin Patch Detection in Real-World Images, Hannes Kruppa, Martin A. Bauer, and Bernt Schiele. To Appear: In DAGM Symposium 2002, Zurich, Switzerland, September 2002. pdf

Abstract: While human skin is relatively easy to detect in controlled environments, detection in uncontrolled settings such as in consumer digital photographs is generally hard. Algorithms need to robustly deal with variations in lighting, color resolution, and imaging noise. This paper proposes a simple generative skin patch model combining shape and color information. The model is parametric and represents the spatial arrangement of skin pixels as compact elliptical regions. Its parameters are estimated by maximizing the mutual information between the model-generated skin pixel distribution and the distribution of skin color as observed in the image. The core of this work is an empirical evaluation on a database of 653 consumer digital photographs. In addition, we investigate the potential of combining our skin detector with state-of-the-art appearance-based face detectors.

Saliency of Interest Points under Scale Changes, Daniela Hall, Bastian Leibe, and Bernt Schiele. To Appear: In British Machine Vision Conference 2002, Cardiff, Wales, September 2002. pdf

Abstract: Interest point detectors are commonly employed to reduce the amount of data to be processed. The ideal interest point detector would robustly select those features which are most appropriate or salient for the application and data at hand. There is however a tradeoff between the robustness and the discriminance of the selected features. Whereas robustness in terms of repeatability is relatively well explored, the discriminance of interest points is rarely discussed. This paper formalizes the notion of saliency and evaluates three state-of-the-art interest point detectors with respect to their capability of selecting salient image features in two recognition settings.

On Performance Prediction and Optimization for Image Retrieval, Julia Vogel and Bernt Schiele. In European Conference on Computer Vision ECCV 2002, Kopenhagen, Denmark, May 2002. pdf ps

Abstract: In content-based image retrieval (CBIR) performance characterization is easily being neglected. A major difficulty lies in the fact that ground truth and the definition of benchmarks are extremely user and application dependent. This paper proposes a two-stage CBIR framework which allows to predict the behavior of the retrieval system as well as to optimize its performance. In particular, it is possible to maximize precision, recall, or jointly precision and recall. The framework is based on the detection of high-level concepts in images. These concepts correspond to vocabulary users can query the database with. Performance optimization is carried out on the basis of the user query, the performance of the concept detectors, and an estimated distribution of the concepts in the database. The optimization is transparent to the user and leads to a set of internal parameters that optimize the succeeding retrieval. Depending only on the query and the desired concept, precision and recall of the retrieval can be increased by up to 40%. The paper discusses the theoretical and empirical results of the optimization as well as its dependency on the estimate of the concept distribution.

Smart CAPs for Smart Its - Context Detection for Mobile Users. Florian Michahelles, Michael Samulowitz, Personal and Ubiquitous Computing, Volume 6, Issue 4 (2002), pp 269-275, Springer, London, September 2002. [Link.Springer.pdf]

Abstract: Context detection for mobile users plays a major role for enabling novel, human-centric interfaces. For this, we introduce a context detection scheme applicable in a self-organized sensor network, which is formed of disseminated, computer empowered sensors, referred to as Smart-Its [1]. Context-detection takes place without requiring any central point of control, and supports push as well as pull modes. Our solution is based on an in-network composition approach relying on so-called smart Context-Aware Packets (sCAPs). These packets act as a uniform interchange format, and allow single sensors to share sensed data and to cooperate to build up a meaningful context model from manifold inputs. sCAPs travel through the sensor network governed by an enclosed retrieving plan, specifying which sensors to visit in order to gain a specific piece of context information. For enhanced flexibility, the retrieving plan itself may be dynamically altered in accordance with past sensor readings.

How Many Classifiers Do I Need?, Bernt Schiele. To appear in International Conference on Pattern Recognition ICPR 2002, Quebec, Canada, August 2002. pdf

Abstract: Combining multiple classifiers promises to increase performance and robustness of a classification task. Currently however, the understanding which combination scheme should be used and the ability to quantify the expected benefit is inadequate. This paper attempts to quantify the performance and robustness gain for different combination schemes and for two classifier types. The results of the paper indicate that the combination of a small number of classifiers may already result in a substantial performance gain. Also, the increase in robustness can be substantial by combining an adequate number of classifiers.

Analyzing Non-Negative Matrix Factorization for Image Classification, David Guillamet, Bernt Schiele, and Jordi Vitri. To appear in International Conference on Pattern Recognition ICPR 2002, Quebec, Canada, August 2002. pdf

Abstract: The Non-negative Matrix Factorization technique (NMF) has been recently proposed for dimensionality reduction. NMF is capable to produce a region- or partbased representation of objects and images. This paper experimentally compares NMF to Principal Component Analysis (PCA) in the context of image patch classification. A first finding is that the two techniques are complementary and that their respective performance is correlated to the with-in class scatter. This paper also analyses different techniques to combine these complementary methods. In the first combination scheme the best technique for each class is chosen and the results are merged. The second combination scheme builds a hierarchy of classifiers where again for each classification task the best technique is chosen. Additionally, incorporation of the classification results of neighboring image patches further improves the overall results.

Detecting Context in Distributed Sensor Networks by Using Smart Context-Aware Packets., Florian Michahelles, Michael Samulowitz and Bernt Schiele. In International Conference on Architecture of Computing Systems (ARCS), 2002, Karlsruhe, Germany, April 2002. pdf

Abstract: Context modeling and detection will play a major role for pervasive computing. This paper proposes an approach to reveal the user's context in a self-organized sensor network without a central point of control. A uniform communication scheme, referred to as Smart Context-Aware Packet's (sCAP's), allows single sensors to share sensed data and to cooperate in order to build a meaningful context model from manifold inputs. In this approach, sCAP's are injected into the sensor network by the context inquirer. In particular, sCAP's contain a retrieving plan specifying which types of sensors should be visited to obtain the desired context information. The paper concentrates on the routing concepts which allow to deal with breakdowns of sensor nodes and continuous changes of the network topology.

3D Object Recognition from Range Images using Local Feature HistogramsGuenter Hetzel, Bastian Leibe, Paul Levi, and Bernt Schiele, To appear in: International Conference on Computer Vision and Pattern Recognition CVPR 2001, Kauai Island, Hawaii, December 2001. cvpr01.pdf

Abstract: This paper explores a view-based approach to recognize free-form objects in range images. We are using a set of local features that are easy to calculate and robust to partial occlusions. By combining those features in a multidimensional histogram, we can obtain highly discriminant classifiers without the need for segmentation. Recognition is performed using either histogram matching or a probabilistic recognition algorithm. We compare the performance of both methods in the presence of occlusions and test the system on a database of almost 2000 full-sphere views of 30 free-form objects. The system achieves a recognition accuracy above 93% on ideal images, and of 89% with 20% occlusion.

Beyond Position Awareness Stavros Antifakos and Bernt Schiele, In: Workshop on Location Modeling for Ubiquitous Computing at UbiComp 2001, Atlanta, October 2001. uc_location01.pdf

Abstract: Location models are merely based on positional information. Using wireless sensor networks however allows to extract information which can be related to different levels of semantic proximity of different devices. Based on this observation the paper proposes a semantic proximity hierarchy based on a wireless sensor network. In this paper we argue that the proposed proximity hierarchy adds a new and complementary dimension to pure positional location information. The paper discusses the proposed hierarchy, gives application examples and some preliminary experimental results.

Performance Prediction for Vocabulary-Supported Image Retrieval Julia Vogel and Bernt Schiele, In: International Conference on Image Processing ICIP 2001, Thessaloniki, Greece, October 2001. icip01.pdf.gz

Abstract: The majority of today's content based image retrieval systems rely on low-level image descriptors which limit their capability to support meaningful interactions with the users. Even though relevance feedback helps, most of the current interaction paradigms are far from the semantic representations which most people use to categorize and describe image content. Therefore we propose a concept called ``vocabulary-supported image retrieval'' which aims to enable the user to access an image database in a more natural way. In particular this paper develops a technique to predict the system's performance with respect to the user query. This allows the system to translate the user query into an internal query which may satisfy predefined criteria such as precision and recall rates. In addition, given the performance parameters of the system's sub-components, the feasibility and the success of the retrieval process can be evaluated beforehand and optimized dynamically online.

Smart-Its Friends: A Technique for Users to Easily Establish Connections between Smart Artefacts Lars Erik Holmquist, Friedemann Mattern, Bernt Schiele, Petteri Alahuhta, Michael Beigl, Hans-W. Gellersen, To appear in: Ubicomp 2001, Atlanta, Gorgia, USA, October 2001.

Abstract: Ubiquitous computing is associated with a vision of everything being connected to everything. However, for successful applications to emerge, it will not be the quantity but the quality and usefulness of connections that will matter. Our concern is how qualitative relations and more selective connections can be established between smart artefacts, and how users can retain control over artefact interconnection. We propose context proximity for selective artefact communication, using the context of artefacts for matchmaking. We further suggest to empower users with simple but effective means to impose the same context on a number of artefacts. To prove our point we have implemented Smart-Its Friends, small embedded devices that become connected when a user holds them together and shakes them.

Smart CAPs for Smart Its - Context Detection for Mobile Users. Florian Michahelles, Michael Samulowitz, In Third International Workshop on Human Computer Interaction with Mobile Devices (MobileHCI), at IHM-HCI 2001, Lille, France, September 2001. [Poster, Draft, Journal]

Abstract: Context detection for mobile users plays a major role for enabling novel, human-centric interfaces. For this, we introduce a context detection scheme for disseminated, computer empowered sensors, referred to as Smart-Its [7]. Context-detection takes place without requiring any central point of control, and supports push as well as pull modes. Our solution is based on an in-network composition approach relying on so-called smart context-aware packets (sCAPs). sCAPs travel thru a sensor network governed by an enclosed retrieving plan, specifying which sensors to visit for gaining a specific piece of context informatio n. For enhanced flexibility, the retrieving plan itself may be dynamically altered in accordance to past sensor readings.

Hierarchical Combination of Object Models using Mutual Information Hannes Kruppa and Bernt Schiele, To appear in: British Machine Vision Conference, BMVC 2001, Manchester, UK, September 2001. bmvc01.pdf

Abstract: Combining different and complementary object models promises to increase the robustness and generality of today's computer vision algorithms. This paper introduces a new method for combining different object models by determining a configuration of the models which maximizes their mutual information. The combination scheme consequently creates a unified hypothesis from multiple object models ``on the fly'' without prior training. To validate the effectiveness of the proposed method, the approach is applied to the detection of faces combining the output of three different models.

Sensory Augmented Computing and its Potential for Human Computer Interaction Bernt Schiele. To appear in HCI International 2001, August 2001.

Abstract: The next generation of computers might be literally wearable. Our vision of such a wearable computing device is an intelligent assistant, which is always with you and helps you to solve your every day tasks. Besides size and power, an important challenge is how to interact with wearable computers. An important aspect of a wearable device is that it can perceive the world from a first-person perspective: a wearable camera can see what you see and a wearable microphone can hear what you hear in order to analyze, model and recognize things and people which are around you. A promising direction for interaction with wearable computers is therefore to make the computers more aware of the situation the user is in and to model the user's context. Sensors, such as cameras, mounted to the user's glasses, can recognize what the user is looking at and might model what the user is doing.

Context-driven Model Switching For Visual Tracking Hannes Kruppa, Martin Spengler and Bernt Schiele, 9th International Symposium on Intelligent Robotic Systems 2001, Toulouse, France, July 2001. sirs01switch.pdf

Abstract: A major challenge for real-world object tracking is the dynamic nature of the environmental conditions with respect to illumination, motion, visibility, etc. For such an environment which may experience drastic changes at any time, integration of multiple and complementary cues promises to increase robustness of visual tracking. Nevertheless, one has to expect that false positive tracking will occur. In order to be able to recover from such tracking failure this paper introduces a novel method for automatically choosing the object model which best fits the current context based on information-theoretic concepts. In order to validate the effectiveness of the proposed model switching, it is integrated into a multi-cue face tracking system and experimentally evaluated.

Towards Robust Multi-cue Integration for Visual Tracking Martin Spengler and Bernt Schiele, In: 9th International Workshop on Computer Vision Systems 2001, p 94-107, Vancouver, Canada, July 2001. pdf, SpringerLink

Abstract: Even though many of today's vision algorithms are very successful, they lack robustness since they are typically limited to a particular situation. In this paper we argue that the principles of sensor and model integration can increase the robustness of today's computer vision systems substantially. As an example multi-cue tracking of faces is discussed. The approach is based on the principles of self-organization of the integration mechanism and self-adaptation of the cue models during tracking. Experiments show that the robustness of simple models is leveraged significantly by sensor and model integration.

Sensory Augmented Computing: Wearing the Museum's Guide Bernt Schiele, T. Jebara, and N. Oliver. In: IEEE Micro, pp 44-52, May/June 2001.

Abstract:A wearable computing device is much more than its desktop counterpart. It is rather like an intelligent assistant that always accompanies you and helps you solve everyday tasks. A key aspect of such devices is that they can operate autonomously and perceive the world as a human user does; without being fed manual input. They augment the user without encumbering him. A smart wearable computer sees what you see and hears what you hear to analyze, recognize and respond to the situations and people you encounter. We describe one incarnation of this smart perceptual remembrance agent, which is equipped with the ability to recognize objects in the user's visual field of view using real-time computer vision. Once an object is recognized, the system displays multimedia information that the user previously identified as being relevant to the object. The computer effectively becomes a tour guide, chiming in with augmented reality to give you reminders as you go about your routine.

Towards Robust Perception and Model Integration Bernt Schiele, Martin Spengler, and Hannes Kruppa. To appear in: Sensor Based Intelligent Robots, Lecture Notes in Computer Science, Springer, 2001.

Abstract: Many of today's vision algorithms are very successful in controlled environments. Real-world environments, however, cannot be controlled and are most often dynamic with respect to illumination changes, motion, occlusions, multiple people, etc. Since most computer vision algorithms are limited to a particular situation they lack robustness in the context of dynamically changing environments. In this paper we argue that the integration of information coming from different visual cues and models is essential to increase robustness as well as generality of computer vision algorithms. Two examples are discussed where robustness of simple models is leveraged by cue and model integration. In the first example mutual information is used as a means to combine different object models for face detection without prior learning. The second example discusses experimental results on multi-cue tracking of faces based on the principles of self-organization of the integration mechanism and self-adaptation of the cue models during tracking.

Vocabulary-Supported Image Retrieval Bernt Schiele and Julia Vogel, In 1st DELOS Workshop on Information Seeking, Searching and Querying in Digital Libraries, Zurich, Switzerland, December 2000. delos00.pdf.gz

Abstract: Today's content-based image retrieval systems (CBIR) mostly rely on a predefined set of low-level image features and incorporate user-interactions using techniques such as relevance feedback. These systems however do not take advantage of the fact that in many applications queries can be formulated using a vocabulary. In this paper we propose a general framework which allows to use vocabulary at several levels. The framework should be seen as an extension of today's CBIR systems enabling the use of vocabulary as well as online learning techniques such as relevance feedback. The image detectors supporting the vocabulary can be either implemented directly or learned offline from examples and user-interactions.

Using Mutual Information to Combine Object Models Hannes Kruppa, Bernt Schiele. In 8th International Symposium on Intelligent Robotic Systems 2000, Reading, UK, July 2000. sirs00mmi.ps.gz

Abstract: This paper introduces a randomized method for combining different object models. By determining a configuration of the models, which maximizes their mutual information, the proposed method creates a unified hypothesis from multiple object models on the fly, without prior training. To validate the effectiveness of the proposed method, experiments are conducted in which human faces are detected and localized in images by combining different face models.

Towards Automatic Extraction and Modelling of Objects from Image Sequences Bernt Schiele. In 8th International Symposium on Intelligent Robotic Systems 2000, Reading, UK, July 2000. sirs00auto.ps.gz

Abstract: This paper exploits a simple but general technique to extract object models from arbitrary image sequences. Such object models can be used to structure and index the image sequence. The algorithm extracts and tracks homogenous regions, which may correspond to objects or object parts. By grouping similar moving regions the algorithm constructs models of potential objects. As such, the approach is model-free in the sense that it does not use a priori models to detect, track and segment objects. On the contrary, the ultimate goal of the approach is to build such models automatically from image sequences.

Recognition without Correspondence using Multidimensional Receptive Field Histograms. Bernt Schiele and James L. Crowley. In International Journal of Computer Vision 36 (1), p 31-50, January 2000 ijcv_schiele.ps.gz

Abstract: The appearance of an object is composed of local structure. This local structure can be described and characterized by a vector of local features measured by local operators such as Gaussian derivatives or Gabor filters. This article presents a technique where appearances of objects are represented by the joint statistics of such local neighborhood operators. As such, this represents a new class of appearance based techniques for computer vision. Based on joint statistics, the paper develops techniques for the identification of multiple objects at arbitrary positions and orientations in a cluttered scene. Experiments show that these techniques can identify over 100 objects in the presence of major occlusions. Most remarkably, the techniques have low complexity and therefore run in real-time.

Probabilistic Object Recognition and Localization. Bernt Schiele and Alex Pentland. In International Conference on Computer Vision, Greece, September 1999