Interactive learning of visually symmetric objects

Li, Wai Ho; Kleeman, Lindsay

doi:10.1109/iros.2009.5354616

Cited by 11 publications

(12 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For learning the appearance and shape of individual objects [15,[25][26][27][28], interaction can also be helpful. Through interaction, a robot can obtain multiple views of a scene (for approaches like [18]) autonomously.…”

Section: B Interactive Perception For Object Segmentationmentioning

confidence: 99%

See 1 more Smart Citation

Probabilistic Segmentation and Targeted Exploration of Objects in Cluttered Environments

2014

View full text Add to dashboard Cite

Abstract-Creating robots that can act autonomously in dynamic, unstructured environments requires dealing with novel objects. Thus, an off-line learning phase is not sufficient for recognizing and manipulating such objects. Rather, an autonomous robot needs to acquire knowledge through its own interaction with its environment, without using heuristics encoding human insights about the domain. Interaction also allows information that is not present in static images of a scene to be elicited. Out of a potentially large set of possible interactions, a robot must select actions that are expected to have the most informative outcomes to learn efficiently. In the proposed bottom-up, probabilistic approach, the robot achieves this goal by quantifying the expected informativeness of its own actions in information-theoretic terms. We use this approach to segment a scene into its constituent objects. We retain a probability distribution over segmentations. We show that this approach is robust in the presence of noise and uncertainty in real-world experiments. Evaluations show that the proposed information-theoretic approach allows a robot to efficiently determine the composite structure of its environment. We also show that our probabilistic model allows straightforward integration of multiple modalities, such as movement data and static scene features. Learned static scene features allow for experience from similar environments to speed up learning for new scenes.

show abstract

Section: B Interactive Perception For Object Segmentationmentioning

confidence: 99%

“…To avoid the segmentation problem, objects can be physically separated from clutter by grasping [22,[27][28][29][30]. Autonomously grasping novel objects, however, is non-trivial by itself and frequently requires knowledge of the object's geometry.…”

Section: B Interactive Perception For Object Segmentationmentioning

confidence: 99%

Probabilistic Segmentation and Targeted Exploration of Objects in Cluttered Environments

2014

View full text Add to dashboard Cite

show abstract

“…Other segmentation methods focused on settings where the robots held an object and used actions to segment it from the background and learn its properties [6,7,12,17,18]. These methods could further inspect objects after they have been segmented and, hence, these approaches would be a good complement to our work.…”

Section: A Interactive Scene Segmentationmentioning

confidence: 99%

“…This bottom-up approach couples perception to physical interactions, such as pushing, grasping, or lifting. Interactive perception allows a robot to learn, for example, the appearance and shape of objects [4][5][6][7][8], their haptic properties [9,10], kinematic structure [2], and how the state of those objects changes as a result of manipulation [4,[10][11][12].…”

Section: Introductionmentioning

confidence: 99%

Maximally informative interaction learning for scene exploration

Hoof

Kroemer

Amor

et al. 2012

2012 IEEE/RSJ International Conference on Intelligent Robots and Systems

View full text Add to dashboard Cite

Abstract-Creating robots that can act autonomously in dynamic, unstructured environments is a major challenge. In such environments, learning to recognize and manipulate novel objects is an important capability. A truly autonomous robot acquires knowledge through interaction with its environment without using heuristics or prior information encoding human domain insights. Static images often provide insufficient information for inferring the relevant properties of the objects in a scene. Hence, a robot needs to explore these objects by interacting with them. However, there may be many exploratory actions possible, and a large portion of these actions may be non-informative. To learn quickly and efficiently, a robot must select actions that are expected to have the most informative outcomes. In the proposed bottom-up approach, the robot achieves this goal by quantifying the expected informativeness of its own actions. We use this approach to segment a scene into its constituent objects as a first step in learning the properties and affordances of objects. Evaluations showed that the proposed information-theoretic approach allows a robot to efficiently infer the composite structure of its environment.

show abstract

“…To resolve this, different strategies have been applied in computational systems, e.g. imposing symmetry constraints on the objects or assuming objects to be placed on a uniformly coloured table top [3,12]. When an object is placed on a multi-coloured table top, with no appearence model associated, the segmented foreground region often covers, not just the object itself, but also parts of the table.…”

Section: Introductionmentioning

confidence: 99%

Active 3D Segmentation through Fixation of Previously Unseen Objects

Björkman¹,

Kragić²

2010

Procedings of the British Machine Vision Conference 2010

View full text Add to dashboard Cite

We present an approach for active segmentation based on integration of several cues. It serves as a framework for generation of object hypotheses of previously unseen objects in natural scenes. Using an approximate Expectation-Maximisation method, the appearance, 3D shape and size of objects are modelled in an iterative manner, with fixation used for unsupervised initialisation. To better cope with situations where an object is hard to segregate from the surface it is placed on, a flat surface model is added to the typical two hypotheses used in classical figure-ground segmentation. The framework is further extended to include modelling over time, in order to exploit temporal consistency for better segmentation and to facilitate tracking.

show abstract

Interactive learning of visually symmetric objects

Cited by 11 publications

References 9 publications

Probabilistic Segmentation and Targeted Exploration of Objects in Cluttered Environments

Probabilistic Segmentation and Targeted Exploration of Objects in Cluttered Environments

Maximally informative interaction learning for scene exploration

Active 3D Segmentation through Fixation of Previously Unseen Objects

Contact Info

Product

Resources

About