Transferring knowledge across robots: A risk sensitive approach

Ciarfuglia, Thomas A.; Valigi, Paolo; Ricci, Elisa

doi:10.1016/j.robot.2014.12.004

Cited by 3 publications

(1 citation statement)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the latest robotics literature a lot of attention has been dedicated to adaptive methods for agents deployed in the real world but trained in simulation environments and with synthetic data produced or collected from free Web resources [21], [22], [23]. Although bridging the so-called reality-gap is important to allow a reduction in the need of manually annotated data, robotic perception provides ample motivations for exploring domain adaptation methods even within real world settings, when changing the visual conditions [24] or when transferring the knowledge acquired by one robot to another [25]. In our work we focus on a single robot that needs to recognize object categories undergoing significant appearance changes due to scaling and translation and we show that the corresponding domain gap can be reduced with a tailored localized adaptive solution based on the identification of domain-invariant image regions.…”

Section: Related Workmentioning

confidence: 99%

Adaptive Deep Learning Through Visual Domain Localization

Angeletti

Caputo

Tommasi

2018

2018 IEEE International Conference on Robotics and Automation (ICRA)

View full text Add to dashboard Cite

A commercial robot, trained by its manufacturer to recognize a predefined number and type of objects, might be used in many settings, that will in general differ in their illumination conditions, background, type and degree of clutter, and so on. Recent computer vision works tackle this generalization issue through domain adaptation methods, assuming as source the visual domain where the system is trained and as target the domain of deployment. All approaches assume to have access to images from all classes of the target during training, an unrealistic condition in robotics applications. We address this issue proposing an algorithm that takes into account the specific needs of robot vision. Our intuition is that the nature of the domain shift experienced mostly in robotics is local. We exploit this through the learning of maps that spatially ground the domain and quantify the degree of shift, embedded into an end-to-end deep domain adaptation architecture. By explicitly localizing the roots of the domain shift we significantly reduce the number of parameters of the architecture to tune, we gain the flexibility necessary to deal with subset of categories in the target domain at training time, and we provide a clear feedback on the rationale behind any classification decision, which can be exploited in human-robot interactions. Experiments on two different settings of the iCub World database confirm the suitability of our method for robot vision.

show abstract