Learning Object Manipulation Skills via Approximate State Estimation from Real Videos - Archive ouverte HAL Access content directly
Conference Papers Year :

Learning Object Manipulation Skills via Approximate State Estimation from Real Videos

(1) , (2) , (2) , (1)
1
2

Abstract

Humans are adept at learning new tasks by watching a few instructional videos. On the other hand, robots that learn new actions either require a lot of effort through trial and error, or use expert demonstrations that are challenging to obtain. In this paper, we explore a method that facilitates learning object manipulation skills directly from videos. Leveraging recent advances in 2D visual recognition and differentiable rendering, we develop an optimization based method to estimate a coarse 3D state representation for the hand and the manipulated object(s) without requiring any supervision. We use these trajectories as dense rewards for an agent that learns to mimic them through reinforcement learning. We evaluate our method on simple single-and two-object actions from the Something-Something dataset. Our approach allows an agent to learn actions from single videos, while watching multiple demonstrations makes the policy more robust. We show that policies learned in a simulated environment can be easily transferred to a real robot.
Fichier principal
Vignette du fichier
2011.06813.pdf (2.18 Mo) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-03017607 , version 1 (21-11-2020)

Identifiers

Cite

Vladimír Petrík, Makarand Tapaswi, Ivan Laptev, Josef Sivic. Learning Object Manipulation Skills via Approximate State Estimation from Real Videos. CoRL 2020 - Conference on Robot Learning, Nov 2020, Virtual, United States. ⟨hal-03017607⟩
44 View
69 Download

Altmetric

Share

Gmail Facebook Twitter LinkedIn More