Skip to content

BerkeleyAutomation/Orienting_Novel_3D_Objects

master
Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?
Code

Latest commit

 

Git stats

Files

Permalink
Failed to load latest commit information.
Type
Name
Latest commit message
Commit time
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Installation:

Create a python3 virtual environment and after enabling the data generation virtual environment, install the necessary requirements as follows:

  1. Run pip install requirements.txt to install all external requirements
  2. Run python setup.py develop to install unsupervised_rbt package

Basic Usage:

You will mostly interact with files in the tools folder, which all have corresponding files in the cfg/tools/ folder which specify paramaters for these files. The controller folder contains scripts for running the simulation experiments to orient objects given a model trained on the self-supervised trask.

Self-Supervised Rotation Prediction Task:

Data Generation: See tools/data_gen_quat.py for generating data for the task. See cfg/tools/data_gen_quat.yaml for config parameters The dataset used for Kit-Net is called 872objv3

Example usage: python tools/data_gen_quat.py {dataset_name}

Training: Make sure to either generate data or use pre-generated data. For training see tools/unsup_rbt_train_quat.py

Example usage: python tools/unsup_rbt_train_quat.py {dataset_name}. Example dataset is 872objv3

Testing: Same as train except with a --test flag.

Example usage: python tools/unsup_rbt_train_quat.py {dataset_name} --test. Example dataset is 872objv3

Prismatic Cavity Task:

Create a dataset with tools/prismatic_cavity.py, then do training and testing as above

Simulation Experiments Controller:

In controller/pyrender_controller.py you can run the orienting objects experiments.

In controller/prism_controller.py you can run the simulation experiments for Kit-Net

About

Orienting Novel 3D Objects Using Self-Supervised Learning of Rotation Transforms

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 4

  •  
  •  
  •  
  •  

Languages