Robot learning of symbol grounding in multiple contexts
Title: Robot learning of symbol grounding in multiple contexts
SNIC Project: Berzelius-2022-158
Project Type: LiU Berzelius
Principal Investigator: Mohamadreza Faridghasemnia <mohamadreza.farid@oru.se>
Affiliation: Örebro universitet
Duration: 2022-08-08 – 2023-03-01
Classification: 10201
Keywords:

Abstract

One of the basic requirements of symbol grounding is that a robot has to recognize objects and their visual properties in an image. I am a WASP AIML Ph.D. student planning to use Berzelius to train a neural network capable of detecting objects and their visual attributes. I plan to use Berzelius to train a model of TensorFlow for a dataset of around 100k images (10gb compressed, almost the same as the visual genome dataset), with about 3000 classes to learn.