Festivus

policy

Transporter Networks

Google Brain / Google Research · TensorFlow

or hover any field below to flag it

Overview

Name
Transporter Networks
Author
Google Brain / Google Research
Framework
TensorFlow
License
apache-2.0
Skill type
manipulation
Evidence level
verified
Task description
Infers spatial displacements from visual input to parameterize pick-and-place actions. Decomposes manipulation into picking attention (where to pick) and placing cross-correlation (where and how to place). Exploits spatial equivariance for extreme sample efficiency -- learns from as few as 10 demos per task.

Spaces

Action space
other · 6-dim · 1Hz
Observation space
  • type: vision
  • · rgbd_topdown (640x480)

Links

HuggingFace repo
null
Paper (arXiv)
https://arxiv.org/abs/2010.14406

Compatible robots

0+1 mentioned but not in catalog yet

No robots list Transporter Networks as compatible yet. Know of one? Flag it above.

Compatible environments

1
tabletop-cleannot in seed

Datasets that reference this policy

0

No datasets reference Transporter Networks yet.