Physical intuition is pivotal for intelligent agents to perform complex
tasks. In this paper we investigate the passive acquisition of an intuitive
understanding of physical principles as well as the active utilisation of this
intuition in the context of generalised object stacking. To this end, we
provide: a simulation-based dataset featuring 20,000 stack configurations
composed of a variety of elementary geometric primitives richly annotated
regarding semantics and structural stability. We train visual classifiers for
binary stability prediction on the ShapeStacks data and scrutinise their
learned physical intuition. Due to the richness of the training data our
approach also generalises favourably to real-world scenarios achieving
state-of-the-art stability prediction on a publicly available benchmark of
block towers. We then leverage the physical intuition learned by our model to
actively construct stable stacks and observe the emergence of an intuitive
notion of stackability - an inherent object affordance - induced by the active
stacking task. Our approach performs well even in challenging conditions where
it considerably exceeds the stack height observed during training or in cases
where initially unstable structures must be stabilised via counterbalancing.
Description
[1804.08018] ShapeStacks: Learning Vision-Based Physical Intuition for Generalised Object Stacking
%0 Generic
%1 groth2018shapestacks
%A Groth, Oliver
%A Fuchs, Fabian B.
%A Posner, Ingmar
%A Vedaldi, Andrea
%D 2018
%K 2018 arxiv computer-vision conference eccv oxford paper robotics
%T ShapeStacks: Learning Vision-Based Physical Intuition for Generalised
Object Stacking
%U http://arxiv.org/abs/1804.08018
%X Physical intuition is pivotal for intelligent agents to perform complex
tasks. In this paper we investigate the passive acquisition of an intuitive
understanding of physical principles as well as the active utilisation of this
intuition in the context of generalised object stacking. To this end, we
provide: a simulation-based dataset featuring 20,000 stack configurations
composed of a variety of elementary geometric primitives richly annotated
regarding semantics and structural stability. We train visual classifiers for
binary stability prediction on the ShapeStacks data and scrutinise their
learned physical intuition. Due to the richness of the training data our
approach also generalises favourably to real-world scenarios achieving
state-of-the-art stability prediction on a publicly available benchmark of
block towers. We then leverage the physical intuition learned by our model to
actively construct stable stacks and observe the emergence of an intuitive
notion of stackability - an inherent object affordance - induced by the active
stacking task. Our approach performs well even in challenging conditions where
it considerably exceeds the stack height observed during training or in cases
where initially unstable structures must be stabilised via counterbalancing.
@misc{groth2018shapestacks,
abstract = {Physical intuition is pivotal for intelligent agents to perform complex
tasks. In this paper we investigate the passive acquisition of an intuitive
understanding of physical principles as well as the active utilisation of this
intuition in the context of generalised object stacking. To this end, we
provide: a simulation-based dataset featuring 20,000 stack configurations
composed of a variety of elementary geometric primitives richly annotated
regarding semantics and structural stability. We train visual classifiers for
binary stability prediction on the ShapeStacks data and scrutinise their
learned physical intuition. Due to the richness of the training data our
approach also generalises favourably to real-world scenarios achieving
state-of-the-art stability prediction on a publicly available benchmark of
block towers. We then leverage the physical intuition learned by our model to
actively construct stable stacks and observe the emergence of an intuitive
notion of stackability - an inherent object affordance - induced by the active
stacking task. Our approach performs well even in challenging conditions where
it considerably exceeds the stack height observed during training or in cases
where initially unstable structures must be stabilised via counterbalancing.},
added-at = {2019-01-15T15:37:13.000+0100},
author = {Groth, Oliver and Fuchs, Fabian B. and Posner, Ingmar and Vedaldi, Andrea},
biburl = {https://www.bibsonomy.org/bibtex/26d1f1bf4b145bdc1bb459dd387439ea6/analyst},
description = {[1804.08018] ShapeStacks: Learning Vision-Based Physical Intuition for Generalised Object Stacking},
interhash = {a964140329abbe9a23c2ee69913a0352},
intrahash = {6d1f1bf4b145bdc1bb459dd387439ea6},
keywords = {2018 arxiv computer-vision conference eccv oxford paper robotics},
note = {cite arxiv:1804.08018Comment: revised version to appear at ECCV 2018},
timestamp = {2019-01-15T15:38:01.000+0100},
title = {ShapeStacks: Learning Vision-Based Physical Intuition for Generalised
Object Stacking},
url = {http://arxiv.org/abs/1804.08018},
year = 2018
}