Object recognition from a single view fails when the available features are not sufficient to determine the identity of a single object, either because of similarity with another object or because of feature corruption due to clutter and occlusion. Active object recognition systems have addressed this problem successfully, but they require complicated systems with adjustable viewpoints that are not always available. In this

Humans learn robust and efficient strategies for visual tasks through interaction with their environment. In contrast, most current computer vision systems have no such learning capabilities. Motivated by insights from psychology and neurobiology, I combine machine learning and computer vision techniques to develop algorithms for visual learning in open-ended tasks. Learning is incremental and makes only weak assumptions about the task environment. I begin

Classification of broad area features in satellite imagery is one of the most important applications of remote sensing. It is often di#cult and time-consuming to develop classifiers by hand, so many researchers have turned to techniques from the fields of statistics and machine learning to automatically generate classifiers. Common techniques include maximum likelihood classifiers, neural networks and genetic algorithms. We present a new system called Afreet, which uses a recently developed machine learning paradigm called Support Vector Machines (SVMs). In contrast to other techniques, SVMs o#er a solid mathematical foundation that provides a probabilistic guarantee on how well the classifier will generalize to unseen data. In addition the SVM training algorithm is guaranteed to converge to the globally optimal SVM classifier, can learn highly non-linear discrimination functions, copes extremely well with high-dimensional feature spaces (such as hyperspectral data), and scales well to large problem sizes. Afreet combines an SVM with a sophisticated spatio-spectral feature construction mechanism that allows it to classify spectrally ambiguous pixels. We demonstrate the e#ectiveness of the system by applying Afreet to several broad area classification problems in remote sensing, and provide a comparison with conventional maximum likelihood classification.

High-level vision systems use object, scene or domain specific knowledge to interpret images. Unfortunately, this knowledge has to be acquired for every domain. This makes it difficult to port systems from one domain to another, and therefore to compare them. Recently, the authors of the ADORE system have claimed that object recognition can be modeled as a Markov decision process, and that domain-specific control strategies can be inferred automatically from training data. In this paper we demonstrate the generality of this approach by porting ADORE to a new domain, where it controls an object recognition system that previously relied on a semantic network.

Abstract. Automated image interpretation is an important task in numerous applications ranging from security systems to natural resource inventorization based on remote-sensing. Recently, a second generation of adaptive machine-learned image interpretation systems have shown expert-level performance in several challenging domains. While demonstrating an unprecedented improvement over hand-engineered and first generation machine-learned systems in terms of cross-domain portability, design-cycle time, and robustness, such systems are still severely limited. This paper inspects the anatomy of the state-of-the-art Multi resolution Adaptive Object Recognition framework (MR ADORE) and presents extensions that aim at removing the last vestiges of human intervention still present in the original design of ADORE. More specifically, feature selection is still a task performed by human domain experts and represents a major stumbling block in the creation process of fully autonomous image interpretation systems. This paper focuses on minimizing such need for human engineering. After discussing experimental results, showing the performance of the framework extensions in the domain of forestry, the paper concludes by outlining autonomous feature extraction methods that may completely remove the need for human expertise in the feature selection process.

MOBSY is a fully integrated autonomous mobile service robot system. It acts

Abstract Automated image interpretation is an important task innumerous applications ranging from security systems to natural resource inventorization based on remote-sensing.Recently, a second generation of adaptive machine-learned image interpretation system (ADORE) has shown expert-level performance in several challenging domains. Its extension, MR ADORE, aims at removing the last vestiges ofhuman intervention still present in the original design of ADORE. Both systems treat the image interpretation pro-cess as a sequential decision making process guided by a machine-learned heuristic value function. This paper em-ploys a new leveraging algorithm for regression (R ESLEV)to improve the learnability of the heuristics in MR ADORE. Experiments show that RESLEV improves the system's per-formance if the base learners are weak. Further analysis discovers the difference between regression and decision-making problems, and suggests an interesting research direction. Keywords: adaptive image interpretation system, leverag-ing for regression, boosting, sequential decision making. 1.

Abstract: One key topic in reinforcement learning is function approximation which is critical for successfully applying reinforcement learning to domains with large state spaces. Unfortunately, function approximation can lead to several problems including the suboptimality of the produced policies and even divergence of learning. Thus, reinforcement learning with function approximation has remained an area of active research. In this paper, we demonstrate that in reinforcement learning, it is helpful for the agent to focus on more important states thereby producing better policies using less computing resources. In particular, the problem of focused learning is investigated formally, and a classification-based reinforcement learning method is considered. We will first define a formal metric of state importance, and then utilize it in reinforcement learning with function approximation. The advantages of focusing attention on important states are supported both theoretically and empirically.

Automated image interpretation and object recognition is an important task in numerous applications ranging from security systems to natural resource inventorization based on remotesensing. Recently, a second generation of adaptive machine-learned image interpretation systems have shown promising performance in several challenging domains. While demonstrating an unprecedented improvement over handengineered or first generation machine learned systems in terms of cross-domain portability, design cycle time, and robustness, such systems are still severely limited. In this paper we pose several open challenges critical to further progress in learning vision systems. The issues are illustrated with recent efforts and examples.

Automated image interpretation is an important task in numerous applications ranging from security systems to natural resource inventorization based on remote-sensing. Recently, a second generation of adaptive machine-learned image interpretation systems have shown expert-level performance in several challenging domains. While demonstrating an unprecedented improvement over hand-engineered and first generation machine-learned systems in terms of cross-domain portability, design-cycle time, and robustness, such systems are still severely limited. This paper reviews the anatomy of the state-of-theart Multi resolution Adaptive Object Recognition framework (MR ADORE) and presents extensions that aim at removing the last vestiges of human intervention still present in the original design of ADORE. More specifically, feature selection is still a task performed by human domain experts thereby prohibiting automatic creation of image interpretation systems. This paper focuses on autonomous feature extraction methods aimed at removing the need for human expertise in the feature selection process.