Contrastive Touch-to-Touch Pretraining

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Rodriguez, Samanta, Dou, Yiming, Bogert, William van den, Oller, Miquel, So, Kevin, Owens, Andrew, Fazeli, Nima
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866916440045518848
author Rodriguez, Samanta
Dou, Yiming
Bogert, William van den
Oller, Miquel
So, Kevin
Owens, Andrew
Fazeli, Nima
author_facet Rodriguez, Samanta
Dou, Yiming
Bogert, William van den
Oller, Miquel
So, Kevin
Owens, Andrew
Fazeli, Nima
contents Today's tactile sensors have a variety of different designs, making it challenging to develop general-purpose methods for processing touch signals. In this paper, we learn a unified representation that captures the shared information between different tactile sensors. Unlike current approaches that focus on reconstruction or task-specific supervision, we leverage contrastive learning to integrate tactile signals from two different sensors into a shared embedding space, using a dataset in which the same objects are probed with multiple sensors. We apply this approach to paired touch signals from GelSlim and Soft Bubble sensors. We show that our learned features provide strong pretraining for downstream pose estimation and classification tasks. We also show that our embedding enables models trained using one touch sensor to be deployed using another without additional training. Project details can be found at https://www.mmintlab.com/research/cttp/.
format Preprint
id arxiv_https___arxiv_org_abs_2410_11834
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Contrastive Touch-to-Touch Pretraining
Rodriguez, Samanta
Dou, Yiming
Bogert, William van den
Oller, Miquel
So, Kevin
Owens, Andrew
Fazeli, Nima
Robotics
Today's tactile sensors have a variety of different designs, making it challenging to develop general-purpose methods for processing touch signals. In this paper, we learn a unified representation that captures the shared information between different tactile sensors. Unlike current approaches that focus on reconstruction or task-specific supervision, we leverage contrastive learning to integrate tactile signals from two different sensors into a shared embedding space, using a dataset in which the same objects are probed with multiple sensors. We apply this approach to paired touch signals from GelSlim and Soft Bubble sensors. We show that our learned features provide strong pretraining for downstream pose estimation and classification tasks. We also show that our embedding enables models trained using one touch sensor to be deployed using another without additional training. Project details can be found at https://www.mmintlab.com/research/cttp/.
title Contrastive Touch-to-Touch Pretraining
topic Robotics
url https://arxiv.org/abs/2410.11834