V0.0.5: R3M Checkpoint

Tag for project R3M: A Universal Visual Representation for Robot Manipulation

R3M studies how to learn generalizable visual representation for robotics from videos of humans and natural language. R3M realizes pre-trained representations (trained on the Ego4D dataset) that are effective in general simulated as well as real-world manipulation tasks. For more details: R3M webpage

Installation

RoboHive uses git submodules to resolve dependencies. Please follow steps exactly as below to install correctly.

1. Clone tag v0.0.5 with pre-populated submodule dependencies

git clone --branch v0.0.5 --recursive https://github.com/vikashplus/robohive.git

2. Install package using pip

cd robohive
pip install -e .

About tag v0.0.5

This checkpoint overrides the default behavior of the environments from a fixed kitchen to one where the kitchen randomly moves around, making the tasks visually more challenging.