Learning viewpoint planning in active recognition on a small sampling budget: a Kriging approach

Joseph Defretin (1, 2) , Julien Marzat (3, 4) , Hélène Piet-Lahanier (3)

1 ONERA - The French Aerospace Lab [Toulouse]
2 CMLA - Centre de Mathématiques et de Leurs Applications
3 ONERA - The French Aerospace Lab [Palaiseau]
4 L2S - Laboratoire des signaux et systèmes

This paper focuses on viewpoint planning for 3D active object recognition. The objective is to design a planning policy into a Q-learning framework with a limited number of samples. Most existing stochastic techniques are therefore inapplicable. We propose to use Kriging and Bayesian Optimization coupled with Q-learning to obtain a computationally-efficient viewpoint-planning design, under a restrictive sampling budget. Experimental results on a representative database, including a comparison with classical approaches, show promising results for this strategy.

Domaines: Optimisation et contrôle [math.OC]; Apprentissage [cs.LG]

Joseph Defretin, Julien Marzat, Hélène Piet-Lahanier. Learning viewpoint planning in active recognition on a small sampling budget: a Kriging approach. 9th IEEE Conference on Machine Learning and Applications, ICMLA 2010, Dec 2010, Washington, D.C., United States. pp.169-174. ⟨hal-00520814⟩