I-POMDP: An infomax model of eye movement

Butko, Nicholas J.; Movellan, Javier R.

doi:10.1109/devlrn.2008.4640819

Cited by 40 publications

(27 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Our robot used active learning exploration strategies to learn to press the buttons. Similar strategies have been formulated in previous work using intrinsically-motivated reinforcement learning [11] [12] [13], and POMDPs [14]. Active learning has been used successfully for grasping objects [15] [16] [17].…”

Section: B Roboticsmentioning

confidence: 99%

Learning to press doorbell buttons

Sukhoy

Sinapov

et al. 2010

2010 IEEE 9th International Conference on Development and Learning

View full text Add to dashboard Cite

Abstract-This paper describes an approach that a robot can use to learn to press doorbell buttons. This approach combines exploratory behaviors with an active learning strategy to enable the robot to learn faster how and where it should press a button in order to trigger the buzzer. The framework was tested with an upper-torso humanoid robot on seven different doorbell buttons. Three different active learning exploration strategies were evaluated: random, stimulus-driven, and uncertaintydriven. The results show that an active learning strategy can significantly speedup the robot's learning progress. Among the three strategies that were evaluated, the uncertainty-driven strategy was the most effective.

show abstract

Section: B Roboticsmentioning

confidence: 99%

Learning to press doorbell buttons

Sukhoy

Sinapov

et al. 2010

2010 IEEE 9th International Conference on Development and Learning

View full text Add to dashboard Cite

show abstract

“…Instead of making random movements, the brain may move the face in more efficient ways to quickly reduce the uncertainly of the internal expression-to-muscle model. Such active exploration may employ information maximization similar to models of human exploratory behavior in eye-movements [13].…”

Section: A Developmental Implicationsmentioning

confidence: 99%

Learning to Make Facial Expressions

Butko

Ruvulo

et al. 2009

2009 IEEE 8th International Conference on Development and Learning

Self Cite

View full text Add to dashboard Cite

Abstract-This paper explores the process of self-guided learning of realistic facial expression production by a robotic head with 31 degrees of freedom. Facial motor parameters were learned using feedback from real-time facial expression recognition from video. The experiments show that these control properties can be learned through an active exploration and feedback loop. The mapping of servos to expressions was learned in under one-hour of training time. We discuss how our work may help illuminate the computational study of how infants learn to make facial expressions.

show abstract

“…Butko & Movellan [4] proposed the I-POMDP framework for modeling visual search. The framework extends the Najemnik & Geisler model by applying long-term POMPDP planning methods.…”

Section: I-pomdp: a Model Of Eye-movementmentioning

confidence: 99%

“…Moreover the optimal search strategy varies in principled ways with the characteristics of the optical device (e.g. eye vs. camera) that is used for searching [4]. While this addressed the first limitation of the Najemik & Geisler model, the second limitation remained unaddressed, i.e, the model was only suitable for a limited class of psychophysical stimuli, namely images that can be described as containing point objects in a field of Gaussian noise.…”

Section: I-pomdp: a Model Of Eye-movementmentioning

confidence: 99%

“…To plan fixations in a way that gathered information close to optimally, we used a policy that was shown to exhibit near-optimal fixation performance for human eyes by Butko & Movellan [4]. This policy biases fixations toward regions of the image where the face is likely to be, and once the location of the face is known with high confidence, the face is always fixated.…”

Section: Default Performancementioning

confidence: 99%

See 1 more Smart Citation

Optimal scanning for faster object detection

Butko

Movellan²

2009

2009 IEEE Conference on Computer Vision and Pattern Recognition

Self Cite

View full text Add to dashboard Cite

Recent years have seen the development of fast and accurate algorithms for detecting objects in images. However, as the size of the scene grows, so do the running-times of these algorithms. If a 128 × 102 pixel image requires 20ms to process, searching for objects in a 1280 × 1024 image will take 2s. This is unsuitable under real-time operating constraints: by the time a frame has been processed, the object may have moved. An analogous problem occurs when controlling robot camera that need to scan scenes in search of target objects. In this paper, we consider a method for improving the run-time of general-purpose object-detection algorithms. Our method is based on a model of visual search in humans, which schedules eye fixations to maximize the long-term information accrued about the location of the target of interest. The approach can be used to drive robot cameras that physically scan scenes or to improve the scanning speed for very large high resolution images. We consider the latter application in this work by simulating a "digital fovea" and sequentially placing it in various regions of an image in a way that maximizes the expected information gain. We evaluate the approach using the OpenCV version of the Viola-Jones face detector. After accounting for all computational overhead introduced by the fixation controller, the approach doubles the speed of the standard Viola-Jones detector at little cost in accuracy.

show abstract

I-POMDP: An infomax model of eye movement

Cited by 40 publications

References 9 publications

Learning to press doorbell buttons

Learning to press doorbell buttons

Learning to Make Facial Expressions

Optimal scanning for faster object detection

Contact Info

Product

Resources

About