Data

Tools

Indexes

Applications

Experiments

Open Source Science

ur5_dataset_builder

https://github.com/chu-2002/ur5_dataset_builder

Science Score: 31.0%

This score indicates how likely this project is to be science-related based on various indicators:

  • ✓
    CITATION.cff file
    Found CITATION.cff file
  • ✓
    codemeta.json file
    Found codemeta.json file
  • ○
    .zenodo.json file
  • ○
    DOI references
  • ○
    Academic publication links
  • ○
    Academic email domains
  • ○
    Institutional organization owner
  • ○
    JOSS paper metadata
  • ○
    Scientific vocabulary similarity
    Low similarity (9.6%) to scientific vocabulary
Last synced: 10 months ago · JSON representation ·

Repository

Basic Info
  • Host: GitHub
  • Owner: CHU-2002
  • Language: Python
  • Default Branch: main
  • Size: 0 Bytes
Statistics
  • Stars: 0
  • Watchers: 1
  • Forks: 0
  • Open Issues: 0
  • Releases: 0
Created over 1 year ago · Last pushed over 1 year ago
Metadata Files
Readme Citation

readme.md

UR5 Robotic Control and Dataset Generation System

This project is used to control the UR5 robot and generate the dataset for the robot operation. Special for the UR5 robot, used for OpenVLA fine-tuning.

Environment Setup

Prerequisites

  • Python 3.10
  • pyrealsense2>=2.55.1.6486
  • opencv-python
  • ur_rtde>=1.5.9
  • numpy==1.24.3
  • pyspacemouse

Installation

```bash

suggest using conda to install the packages

conda create -n ur5controller python=3.10 conda activate ur5controller

Install pyspacemouse

According to the official website: https://github.com/JakubAndrysek/PySpaceMouse
Possible instuctions: https://blog.csdn.net/qq_40081208/article/details/137675822 https://bbs.archlinux.org/viewtopic.php?id=278341

Install required Python packages

pip install -r requirements.txt

```

Project Structure

├── ur5_controller/ │ ├── get_pos_revised_class.py # Main control logic │ ├── vacuum_gripper.py # Gripper control interface │ ├── get_pos_revised_class_moveL.py # Main control logic for moveL │ └── __init__.py # Package exports ├── ur5_robo_dataset_dataset_builder.py # Dataset generation pipeline └── readme.md # This documentation

Basic Usage

Start the robot control script and save the data to the dataset folder. bash ./ur5_controller/get_pos_revised_class_moveL.py clean the dataset folder. bash ./filter_dataset.py build the dataset. bash tfds build

Data Collection Protocol

Raw Dataset Structure

data/ ├── task0/ │ ├── target0.npy │ ├── target1.npy │ ├── target2.npy │ └── ... ├── task1/ │ ├── target0.npy │ ├── target1.npy │ ├── target2.npy │ └── ... ├── task2/ └── ...

Raw Data Fields (per step)

  • RGB images (3x640x480)
  • Depth maps (1x640x480)
  • End-effector pose (6D vector)
  • Joint positions (6D vector)
  • Gripper state (0 or 1)

Player:

uesd to play the raw data

vidio_gen:

used to generate the video from the raw data

Owner

  • Name: YiTong
  • Login: CHU-2002
  • Kind: user

Citation (CITATIONS.bib) 

// TODO(example_dataset): BibTeX citation

GitHub Events

Total
  • Watch event: 2
  • Push event: 2
  • Create event: 2
Last Year
  • Watch event: 2
  • Push event: 2
  • Create event: 2

Dependencies

requirements.txt pypi
  • numpy ==1.24.3
  • opencv-python *
  • pyrealsense2 >=2.55.1.6486
  • ur_rtde >=1.5.9