Mapping Instructions to Actions in 3D Environments with Visual Goal Prediction

Andrew Bennett
Andrew Bennett
Valts Blukis
Valts Blukis
Eyvind Niklasson
Eyvind Niklasson
Max Shatkhin
Max Shatkhin

EMNLP, pp. 2667-2678, 2018.

Cited by: 34|Views70
EI

Abstract:

We propose to decompose instruction execution to goal prediction and action generation. We design a model that maps raw visual observations to goals using LINGUNET, a language-conditioned image generation network, and then generates the actions required to complete them. Our model is trained from demonstration only without external resour...More

Code:

Data:

Your rating :
0

 

Tags
Comments