In human-robot collaboration scenarios, mutual adaptation between the human and robot must occur to ensure high task performance. This requires robotic systems to be capable of reasoning based on a long-term history of interactions. In this paper, we present and evaluate a robot simulation system that facilitates adaptive robot behavior using ontology-based reasoning and behavior trees in an interactive robotic scanning task. A study with 38 participants compares our adaptive system with a static system in team performance and perceived system usability. Our results suggest that use of the adaptive system significantly reduced session time, leading users to perform the task 19.5% faster. Furthermore, participants reported significantly lower fatigue levels, while maintaining the same task performance as those using the static system.

Given a setup with external cameras and a mobile manipulator with an eye-in-hand camera, we address theproblem of computing a sequence of base poses and grasp choices that allows for clearing objects from atable while minimizing the overall execution time. The first step in our approach is to construct a worldmodel, which is generated by an anchoring process, using information from the external cameras. Next, wedeveloped a planning module which – based on the contents of the world model - is able to create a plausibleplan for reaching base positions and suitable grasp choices keeping execution time minimal. Comparing ourapproach to two baseline methods shows that the average execution cost of plans computed by our approach is40% lower than the naive baseline and 33% lower than the heuristic-based baseline. Furthermore, we integrateour approach in a demonstrator, undertaking the full complexity of the problem.

In many applications, a mobile manipulator robot is required to grasp a set of objects distributed in space. This may not be feasible from a single base pose and the robot must plan the sequence of base poses for grasping all objects, minimizing the total navigation and grasping time. This is a Combinatorial Optimization problem that can be solved using exact methods, which provide optimal solutions but are computationally expensive, or approximate methods, which offer computationally efficient but sub-optimal solutions. Recent studies have shown that learning-based methods can solve Combinatorial Optimization problems, providing near-optimal and computationally efficient solutions.In this work, we present BaSeNet - a learning-based approach to plan the sequence of base poses for the robot to grasp all the objects in the scene. We propose a Reinforcement Learning based solution that learns the base poses for grasping individual objects and the sequence in which the objects should be grasped to minimize the total navigation and grasping costs using Layered Learning. As the problem has a varying number of states and actions, we represent states and actions as a graph and use Graph Neural Networks for learning. We show that the proposed method can produce comparable solutions to exact and approximate methods with significantly less computation time. The code and Reinforcement Learning environments will be made available on the project webpage*.

Due to demographic change, health and elderly care systems are facing a shortage of qualified caregivers. This issue can be addressed by introducing welfare robots into people’s homes, hospitals, and care institutions. To provide useful support, such robots must adapt to individual users and smoothly interact with them. From this perspective, we present advances on the development of proactive control for online individual user adaptation in a welfare robot guidance scenario, with the integration of three main modules: 1) navigation control; 2) visual human detection; and 3) temporal error correlation-based neural learning. The proposed control approach can drive a mobile robot to autonomously navigate in relevant indoor environments. At the same time, it can predict human walking speed based on visual information without prior knowledge of personality and preferences (i.e., walking speed). The robot then uses this prediction to continuously adapt its speed to individual users in a proactive online manner. We validate the performance of the proposed proactive robot control in different real-world environments with various users, including an elderly resident of a Danish elderly care center. The results show that the robot successfully and smoothly guided various users of different ages and average walking speeds (e.g., 0.2 m/s, 0.7 m/s, and 1.1 m/s) to target locations over distances of 25–60 m. All in all, this study captures a wide range of research from robot control technology development to technological validity in a relevant environment and system prototype demonstration in an operational environment (i.e., an elderly care center).

The ability to track the 6D pose distribution of an object when a mobile manipulator robot is still approaching the object can enable the robot to pre-plan grasps that combine base and arm motion. However, tracking a 6D object pose distribution from a distance can be challenging due to the limited view of the robot camera. In this work, we present a framework that fuses observations from external stationary cameras with a moving robot camera and sequentially tracks it in time to enable 6D object pose distribution tracking from a distance. We model the object pose posterior as a multi-modal distribution which results in a better performance against uncertainties introduced by large camera-object distance, occlusions and object geometry. We evaluate the proposed framework on a simulated multi-view dataset using objects from the YCB data set. Results show that our framework enables accurate tracking even when the robot camera has poor visibility of the object.

With the emergence of robots being deployed in unstructured environments outside the industrial domain, the importance of robots behaving appropriately in the vicinity of people is becoming more clear. These behaviours are hard to model as they depend on the social context. This context includes among other things where the robot is deployed, how crowded that place is, as well as who are residing in that place. In this paper we extend social space theory with the social context, making them adaptable to the current situation. We implement the social spaces as costmaps used in the standard ROS navigation stack. Our method - Context-Aware Social robot Navigation (CASN) - is tested in the context of people avoidance in social navigation. We compare CASN with the social navigation layer package, which also implements costs based on detected people. We show that by using CASN a mobile robot complies with social conventions in four different navigation scenarios.

In this paper, we present a method for proactive approaching of humans for human-robot cooperative tasks such as a robot serving beverages to people. The proposed method can deal robustly with the uncertainties in the robot’s perception while also ensuring socially acceptable behavior. We use multiple modalities in the form of the robot’s motion, body orientation, speech and gaze to proactively approach humans. Further, we present a behavior tree based control architecture to efficiently integrate these different modalities. The proposed method was successfully integrated and tested on a beverage serving robot. We present the findings of our experiments and discuss possible extensions to address limitations.

Hand disinfection in public spaces is of great importance in preventing infectious diseases. However not everyone sanitizes their hands using hand sanitizer dispensers, and even if they do, many of them don’t rub hands for a long enough time for the sanitizing agent to become most effective. For these reasons we designed a robotic interface for automatic hand sanitizer dispensers to motivate people to disinfect their hands more often and for a longer time. We use interactive elements like speech and gaze communication to achieve this. In our in-the-wild studies we have found that using our system resulted in 21% more hand sanitizations and a much longer hand rubbing time, which in turn leads to better public hygiene and better prevention of infectious diseases.

In this paper, we propose an end-to-end-solution to the problem of multi-object tracking on a mobile robot. The tracking system consists of a process where we project 2D multi-object detections to the robots base frame, using RGB-D sensor data. These detections are then transformed to the map frame using a localization algorithm. This system predicts trajectories of humans and objects in the environment of the robot and can be adapted to work with any detector and track from multiple cameras. The system can then be used to build a temporally consistent costmap to improve navigation strategies.

Vision systems are often used jointly with robotic manipulators to perform automated tasks in industrial applications. Still, the correct set up of such workcells is difficult and requires significant resources. One of the main challenges, when implementing such systems in industrial use cases, is the pose uncertainties presented by the vision system which have to be handled by grasping. In this paper, we present a framework for the design and analysis of optimal gripper finger designs and vision parameters. The proposed framework consists of two parallel methods which rely on vision and grasping simulation to provide an initial estimation of the uncertainty compensation capabilities of the designs. In case the compensation is not feasible with the initial design, an optimization process is introduced, to select the optimal pose estimation parameters and finger designs for the presented task. The proposed framework was evaluated in dynamic simulation and implemented in a real industrial use case.

We present a three-level cognitive system in a learning by demonstration context. The system allows for learning and transfer on the sensorimotor level as well as the planning level. The fundamentally different data structures associated with these two levels are connected by an efficient mid-level representation based on so-called “semantic event chains.” We describe details of the representations and quantify the effect of the associated learning procedures for each level under different amounts of noise. Moreover, we demonstrate the performance of the overall system by three demonstrations that have been performed at a project review. The described system has a technical readiness level (TRL) of 4, which in an ongoing follow-up project will be raised to TRL 6.

Gripper design is nowadays an area of ongoing research activity. The problem of creating a generic and automated gripper design approach tailored for a specific task is still far from solved. In this paper, we propose a new method of generating finger cut-outs aimed at simplifying the design process of doing so. This method takes root in the idea of using the imprint to produce the finger geometry. We furthermore provide a verification of our newly introduced imprinting method and a comparison to the previously introduced parametrized geometry method. This verification is done through a set of grasping experiments performed in simulation on two objects with geometry features based on those found in industrial setting.

A general requirement for any service robot is to be flexible and capable of processing uncertainties, thus making it adaptable for multiple tasks. As a result, learning the appropriate action parameters for a specific action is a crucial task. The method presented in this paper is an incremental statistical learning method that takes into consideration the uncertainties and the contact forces to find the optimal parameter sets. The method is inspired by solutions available in industrial robotics and it uses a dynamic simulator and Kernel Density Estimation in order to find the parameter sets that lead to a successful Peg-in-Hole action. The solution obtained in the simulation is successfully tested on a real service robot.

This paper presents a new large scale dataset targeting evaluation of local shape descriptors and 3d object recognition algorithms. The dataset consists of point clouds and triangulated meshes from 292 physical scenes taken from 11 different views, a total of approximately 3204 views. Each of the physical scenes contain 10 occluded objects resulting in a dataset with 32040 unique object poses and 45 different object models. The 45 object models are full 360 degree models which are scanned with a high precision structured light scanner and a turntable. All the included objects belong to different geometric groups, concave, convex, cylindrical and flat 3D object models. The object models have varying amount of local geometric features to challenge existing local shape feature descriptors in terms of descriptiveness and robustness. The dataset is validated in a benchmark which evaluates the matching performance of 7 different state-of-the-art local shape descriptors. Further, we validate the dataset in a 3D object recognition pipeline. Our benchmark shows as expected that local shape feature descriptors without any global point relation across the surface have a poor matching performance with flat and cylindrical objects. It is our objective that this dataset contributes to the future development of next generation of 3D object recognition algorithms. The dataset is public available at http://roboimagedata.compute.dtu.dk/.

Programming by Demonstration (PbD) is an efficient way for non-experts to teach new skills to a robot. PbD can be carried out in different ways, for instance, by kinesthetic guidance, teleoperation or by using external controls. In this paper, we compare these three ways of controlling a robot in terms of efficiency, effectiveness (success and error rate) and usability. In an industrial assembly scenario, 51 participants carried out peg-in-hole tasks using one of the three control modalities. The results show that kinesthetic guidance produces the best results. In order to test whether the problems during teleoperation are due to the fact that users cannot, like in kinesthetic guidance, switch between control points using traditional teleoperation devices, we designed a new device that allows users to switch between controls for large and small movements. A user study with 15 participants shows that the novel teleoperation device yields almost as good results as kinesthetic guidance.

This paper discusses various optimization schemes for partly stochastic and bound optimization, particular with application to solve robotic optimization problems, where robustness of the solutions is crucial. The use case revolves around grasping and manipulation of deformable objects. These kinds of tasks are difficult to tune to a satisfactory extent by expert users, and therefore optimization by simulation is a good alternative to achieve satisfactory solutions. In order to apply the optimization, a dynamic simulation framework has been used to model the performance of a given solution for the task. The solutions are parameterized in terms of the robot motion and the gripper configuration, and after each simulation various objective scores are determined and combined. This enables the use of various optimization strategies. Based on visual inspection of the most robust solution found, it is determined that 50 out of 50 simulations, with different meat properties, produce satisfactory manipulations.

In this paper, we present an object categorization system capable of assigning multiple and related categories for novel objects using multi-label learning. In this system, objects are described using global geometric relations of 3D features. We propose using the Joint SVM method for learning and we investigate the extraction of hierarchical clusters as a higher-level description of objects to assist the learning. We make comparisons with other multi-label learning approaches as well as single-label approaches (including a state-of-the-art methods using different object descriptors). The experiments are carried out on a dataset of 100 objects belonging to 13 visual and action-related categories. The results indicate that multi-label methods are able to identify the relation between the dependent categories and hence perform categorization accordingly. It is also found that extracting hierarchical clusters does not lead to gain in the system's performance. The results also show that using histograms of global relations to describe objects leads to fast learning in terms of the number of samples required for training.

We present a method for finding correspondence between 3D models. From an initial set of feature correspondences, our method uses a fast voting scheme to separate the inliers from the outliers. The novelty of our method lies in the use of a combination of local and global constraints to determine if a vote should be cast. On a local scale, we use simple, low-level geometric invariants. On a global scale, we apply covariant constraints for finding compatible correspondences. We guide the sampling for collecting voters by downward dependencies on previous voting stages. All of this together results in an accurate matching procedure. We evaluate our algorithm by controlled and comparative testing on different datasets, giving superior performance compared to state of the art methods. In a final experiment, we apply our method for 3D object detection, showing potential use of our method within higher-level vision.

This paper presents a fully calibrated extended geometric approach for gaze estimation in three dimensions (3D). The methodology is based on a geometric approach utilising a fully calibrated binocular setup constructed as a head-mounted system. The approach is based on utilisation of two ordinary web-cameras for each eye and 6D magnetic sensors allowing free head movements in 3D. Evaluation of initial experiments indicate comparable results to current state-of-the-art on estimating gaze in 3D. Initial results show an RMS error of 39-50 mm in the depth dimension and even smaller in the horizontal and vertical dimensions regarding fixations. However, even though the workspace is limited, the fact that the system is designed as a head-mounted device, the workspace volume is relatively positioned to the pose of the device. Hence gaze can be estimated in 3D with relatively free head-movements with external reference to a world coordinate system and is therefore offering flexibility and movability within certain constraints.

In this paper we present an object detection method that uses edge categorisation in combination with a local multi-modal histogram descriptor, all based on RGB-D data. Our target application is robust detection and pose estimation of known objects. We propose to apply a recently introduced edge categorisation algorithm for describing objects in terms of its different edge types. Relying on edge information allow our system to deal with objects with little or no texture or surface variation. We show that edge categorisation improves matching performance due to the higher level of discrimination, which is made possible by the explicit use of edge categories in the feature descriptor. We quantitatively compare our approach with the state-of-the-art template based Linemod method, which also provides an effective way of dealing with texture-less objects, tests were performed on our own object dataset. Our results show that detection based on edge local multi-modal histogram descriptor outperforms Linemod with a significantly smaller amount of templates.

Effective robot manipulation requires a vision system which can extract features of the environment which determine what manipulation actions are possible. There is existing work in this direction under the broad banner of recognising “affordances”. We are particularly interested in possibilities for actions afforded by relationships among pairs of objects. For example if an object is “inside” another or “on top” of another. For this there is a need for a vision system which can recognise such relationships in a scene. We use an approach in which a vision system first segments an image, and then considers a pair of objects to determine their physical relationship. The system extracts surface patches for each object in the segmented image, and then compiles various histograms from looking at relationships between the surface patches of one object and those of the other object. From these histograms a classifier is trained to recognise the relationship between a pair of objects. Our results identify the most promising ways to construct histograms in order to permit classification of physical relationships with high accuracy. This work is important for manipulator robots who may be presented with novel scenes and must identify the salient physical relationships in order to plan manipulation activities.

This paper describes an adaptable system which is able to perform manipulation operations (such as Peg-in-Hole or Laying-Down actions) with flexible objects. As such objects easily change their shape significantly during the execution of an action, traditional strategies, e.g, for solve path-planning problems, are often not applicable. It is therefore required to integrate visual tracking and shape reconstruction with a physical modeling of the materials and their deformations as well as action learning techniques. All these different submodules have been integrated into a demonstration platform, operating in real-time. Simulations have been used to bootstrap the learning of optimal actions, which are subsequently improved through real-world executions. To achieve reproducible results, we demonstrate this for casted silicone test objects of regular shape. Note to Practitioners - The aim of this work was to facilitate the setup of robot-based automation of delicate handling of flexible objects consisting of a uniform material. As examples, we have considered how to optimally maneuver flexible objects through a hole without colliding and how to place flexible objects on a flat surface with minimal introduction of internal stresses in the object. Given the material properties of the object, we have demonstrated in these two applications how the system can be programmed with minimal requirements of human intervention. Rather than being an integrated system with the drawbacks in terms of lacking flexibility, our system should be viewed as a library of new technologies that have been proven to work in close to industrial conditions. As a rather basic, but necessary part, we provide a technology for determining the shape of the object when passing on, e.g., a conveyor belt prior to being handled. The main technologies applicable for the manipulated objects are: A method for real-time tracking of the flexible objects during manipulation, a method for model-based offline prediction of the static deformation of grasped, flexible objects and, finally, a method for optimizing specific tasks based on both simulated and real-world executions.

We address the problem of estimating the alignment pose between two models using structure-specific local descriptors. Our descriptors are generated using a combination of 2D image data and 3D contextual shape data, resulting in a set of semi-local descriptors containing rich appearance and shape information for both edge and texture structures. This is achieved by defining feature space relations which describe the neighborhood of a descriptor. By quantitative evaluations, we show that our descriptors provide high discriminative power compared to state of the art approaches. In addition, we show how to utilize this for the estimation of the alignment pose between two point sets. We present experiments both in controlled and real-life scenarios to validate our approach.

We present an object recognition system coding shape by view-point invariant geometric relations and appearance. In our intelligent work-cell, the system can observe the work space of the robot by 3 pairs of Kinect and stereo cameras allowing for reliable and complete object information. We show that in such a set-up we can achieve high performance already with a low number of training samples. We show this by training the system to classify 56 objects using Random Forest algorithm. This indicates that our approach can be used in contexts such as assembly manipulation which require high reliability of object recognition.

Computational modeling of the primate visual system yields insights of potential relevance to some of the challenges that computer vision is facing, such as object recognition and categorization, motion detection and activity recognition, or vision-based navigation and manipulation. This paper reviews some functional principles and structures that are generally thought to underlie the primate visual cortex, and attempts to extract biological principles that could further advance computer vision research. Organized for a computer vision audience, we present functional principles of the processing hierarchies present in the primate visual system considering recent discoveries in neurophysiology. The hierarchical processing in the primate visual system is characterized by a sequence of different levels of processing (on the order of 10) that constitute a deep hierarchy in contrast to the flat vision architectures predominantly used in today's mainstream computer vision. We hope that the functional description of the deep hierarchies realized in the primate visual system provides valuable insights for the design of computer vision algorithms, fostering increasingly productive interaction between biological and computer vision research.

In this paper, we present what appears to be the first studies of how to apply learning methods for improving the grasp success probability in industrial bin picking. Our study comprises experiments with both a pneumatic parallel gripper and a suction cup. The baseline is a prioritized list of grasps that have been chosen manually by an experienced engineer. We discuss generally the probability space for success probability in bin picking and we provide suggestions for robust success probability estimates for difference sizes of experimental sets. By performing grasps equivalent to one or two days in production, we show that the success probabilities can be significantly improved by the proposed learning procedure.

For some time now machine learning methods have been widely used in perception for autonomous robots. While there have been many results describing the performance of machine learning techniques with regards to their accuracy or convergence rates, relatively little work has been done on developing theoretical performance guarantees about their stability and robustness. As a result, many machine learning techniques are still limited to being used in situations where safety and robustness are not critical for success. One way to overcome this difficulty is by using reachability analysis, which can be used to compute regions of the state space, known as reachable sets, from which the system can be guaranteed to remain safe over some time horizon regardless of the disturbances. In this paper we show how reachability analysis can be combined with machine learning in a scenario in which an aerial robot is attempting to learn the dynamics of a ground vehicle using a camera with a limited field of view. The resulting simulation data shows that by combining these two paradigms, one can create robotic systems that feature the best qualities of each, namely high performance and guaranteed safety.

To offer increased security and comfort, advanced driver-assistance systems (ADASs) should consider individual driving styles. Here, we present a system that learns a human's basic driving behavior and demonstrate its use as ADAS by issuing alerts when detecting inconsistent driving behavior. In contrast to much other work in this area, which is based on or obtained from simulation, our system is implemented as a multithreaded parallel central processing unit (CPU)/graphics processing unit (GPU) architecture in a real car and trained with real driving data to generate steering and acceleration control for road following. It also implements a method for detecting independently moving objects (IMOs) for spotting obstacles. Both learning and IMO detection algorithms are data driven and thus improve above the limitations of model-based approaches. The system's ability to imitate the teacher's behavior is analyzed on known and unknown streets, and results suggest its use for steering assistance but limit the use of the acceleration signal to curve negotiation. We propose that this ability to adapt to the driver can lead to better acceptance of ADAS, which is an important sales argument.

We present a method for learning object grasp affordance models in 3D from experience, and demonstrate its applicability through extensive testing and evaluation on a realistic and largely autonomous platform. Grasp affordance refers here to relative object-gripper configurations that yield stable grasps. These affordances are represented probabilistically with grasp densities, which correspond to continuous density functions defined on the space of 6D gripper poses. A grasp density characterizes an object's grasp affordance; densities are linked to visual stimuli through registration with a visual model of the object they characterize. We explore a batch-oriented, experience-based learning paradigm where grasps sampled randomly from a density are performed, and an importance-sampling algorithm learns a refined density from the outcomes of these experiences. The first such learning cycle is bootstrapped with a grasp density formed from visual cues. We show that the robot effectively applies its experience by downweighting poor grasp solutions, which results in increased success rates at subsequent learning cycles. We also present success rates in a practical scenario where a robot needs to repeatedly grasp an object lying in an arbitrary pose, where each pose imposes a specific reaching constraint, and thus forces the robot to make use of the entire grasp density to select the most promising achievable grasp.

In this paper we present a scheme for accumulating local visual information in 3D, under known motion. Information about the object’s 3D shape is provided by reconstructing local contour descriptors. This shape information is accumulated over time in three ways: 1) disambiguation: erroneous stereo correspondences that are unsuccessfully tracked are discarded. We make use of aspect cues to increase the data association selectivity. 2) correction: the full pose of the reconstructed features is corrected over time using an Kalman Filter approach. 3) completeness: multiple 2 1/2D representations become merged, constructing a full 3D representation of the object. The described system is evaluated quantitatively on three different scenarios.

Visually extracted 2D and 3D information have their own advantages and disadvantages that complement each other. Therefore, it is important to be able to switch between the different dimensions according to the requirements of the problem and use them together to combine the reliability of 2D information with the richness of 3D information. In this article, we use 2D and 3D information in a feature-based vision system and demonstrate their complementary properties on different applications (namely: depth prediction, scene interpretation, grasping from vision and object learning).

For the analysis of images, a deeper understanding of their intrinsic structure is required. This has been obtained for 2D images by means of statistical analysis [15, 18]. Here, we analyze the relation between local image structures (i.e., homogeneous, edge-like, corner-like or texturelike structures) and the underlying local 3D structure, represented in terms of continuous surfaces and different kinds of 3D discontinuities, using 3D range data with the true color information. We find that homogeneous image patches correspond to continuous surfaces, and discontinuities are mainly formed by edge-like or corner-like structures. The results are discussed with regard to existing and potential computer vision applications and the assumptions made by these applications.

The intrinsic dimension (see, e.g., [29, 11]) has proven to be a suitable descriptor to distinguish between different kind of image structures such as edges, junctions or homogeneous image patches. In this paper, we will show that the intrinsic dimension is spanned by two axes: one axis represents the variance of the spectral energy and one represents the a weighted variance in orientation. Moreover, we will show in section that the topological structure of instrinsic dimension has the form of a triangle. We will review diverse definitions of intrinsic dimension and we will show that they can be subsumed within the above mentioned scheme. We will then give a concrete continous definition of intrinsic dimension that realizes its triangular structure.

We introduce a compact coding of image information which explicitly sep- arates visual information into geometric information (orientation) and struc- tural information (phase and colour) and temporal information (optic flow). We investigate the importance of these visual attributes for stereo match- ing on a large data set. From these investigation we can conclude that it is the combination of different attributes that gives the best results. Concrete weights for the relative importance of different visual attributes are statisti- cally determined.

In this article, a biologically plausible and efficient object recognition system (called ORASSYLL) is introduced, based on a set of a priori constraints motivated by findings of developmental psychology and neurophysiology. These constraints are concerned with the organization of the input in local and corresponding entities, the interpretation of the input by its transformation in a highly structured feature space, and the evaluation of features extracted from an image sequence by statistical evaluation criteria. In the context of the bias-variance dilemma, the functional role of a priori knowledge within ORASSYLL is discussed. In contrast to systems in which object representations are defined manually,the introduced constraints allow an autonomous learning from complex scenes.

We introduce an object recognition system in which objects are represented as a sparse and spatially organized set of local (bent) line segments. The line segments correspond to binarized Gabor wavelets or banana wavelets, which are bent and stretched Gabor wavelets. These features can be metrically organized, the metric enables an efficient learning of object representations. Learning can be performed autonomously by utilizing motor-- controlled feedback. The learned representation are used for fast and efficient localization and discrimination of objects in complex scenes. 1 Introduction In this paper we describe a novel object recognition system called ORASSYLL (Object Recognition with Autonomously learned and Sparse SYmbolic representations based on Local Line detectors). In ORASSYLL representations of object classes can be learned autonomously. The learned representations are used for a fast and efficient location and identification of objects in complicated scenes. Lea...

We present a system for recognizing human faces from single images out of a large database containing one image per person. Faces are represented by labeled graphs, based on a Gabor wavelet transform. Image graphs of new faces are extracted by an elastic graph matching process and can be compared by a simple similarity function. The system differs from Lades et al. (1993) in three respects. Phase information is used for accurate node positioning. Object-adapted graphs are used to handle large rotations in depth. Image graph extraction is based on a novel data structure, the bunch graph, which is constructed from a small set of sample image graphs.

We introduce a learning algorithm for the weights in a very common class of discrimination functions usually called "weighted average." The learning algorithm can reduce the number of free variables by simple but effective a priori criteria about significant features. Here we apply our algorithm to three tasks of different dimensionality all concerned with face recognition.

We present a system for recognizing human faces from single images out of a large database containing one image per person. Faces are represented by labeled graphs, based on a Gabor wavelet transform. Image graphs of new faces are extracted by an elastic graph matching process and can be compared by a simple similarity function. The system differs from the preceding one (Lades et al., 1993) in three respects. Phase information is used for accurate node positioning. Object-adapted graphs are used to handle large rotations in depth. Image graph extraction is based on a novel data structure, the bunch graph, which is constructed from a small get of sample image graphs.

We present a new system for the automatic determination of the position, size and pose of the head of a human figure in a camera image. The system is an extension of the well--known face recognition system [WFK] to pose estimation. The pose estimation system is characterized by a certain reliability and speed. We improve this performance and speed with the help of statistical estimation methods. In order to make these applicable, we reduce the originally very high dimensionality of our system with the help of a number of a priori principles. 1 Introduction In this paper we deal with two problems. Firstly, we describe a pose estimation algorithm based on Elastic Graph Matching (EGM) [LVB, WFK]. The algorithm is an extension of the face representation introduced in [WFK] to the problem of pose estimation, in [WFK] the poses of faces is assumed to be known. Secondly, we improve the performance and speed of the pose estimation algorithm by learning. This learning algorithm can be ...