# VersatileGrasping

**Repository Path**: goodbo/VersatileGrasping

## Basic Information

- **Project Name**: VersatileGrasping
- **Description**: No description available
- **Primary Language**: Python
- **License**: MIT
- **Default Branch**: main
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2024-12-24
- **Last Updated**: 2024-12-24

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# Introduction
### NOTE: In this repository, we only provide the code for the GA-CNN training with a mini dataset.

## Abstract

<img src="tutorials_imgs/introduction_01.png" width="100%" alt="abstract">
<img src="tutorials_imgs/01_flowchart_large.svg" width="100%" alt="flowchart">

## Extra explanation
1) This repository has been tested on Ubuntu 16.0 (python 3.6) and Ubuntu 20.0 (python 3.8), and the following tutorial is based on Ubuntu 20.0 (python 3.8).


# Video demo
1. https://youtu.be/aPh1wOYH1Pw
2. https://1drv.ms/v/s!Aok6lAYtb5vYzSLKD_62hTI4DCRa?e=WLShMW (backup)

# Training a neural network
1. Here we provide a brief training demo based on a mini dataset. Please download the mini dataset [here](https://drive.google.com/file/d/1fJBxswzjU5H4lqjVxG3UMEkH3Qm9gaUu/view?usp=sharing).

2. Unzip the mini dataset and copy them into the path `$HOME/$PATH OF YOUR REPOSITORY$/dataset`.  
    <img src="tutorials_imgs/dataset_example.png" width="50%" alt="dataset">  
    
3. Launch the visualization webpage:
    ```bash
    cd $HOME/$PATH OF YOUR REPOSITORY$/NeuralNetwork/data
    python -m visdom.server -port 8031 -env_path ~/$PATH OF YOUR REPOSITORY$/NeuralNetwork/data
    ```
4. Open your web browser, and visit the webpage below to monitor the training progress:
    ```bash
    http://localhost:8031/
    ```

5. Start training:
    ```bash
    cd $HOME/$PATH OF YOUR REPOSITORY$/NeuralNetwork/data
    python train_DexVacuum_Linr_80.py
    ```

### Extra tips for neural network training

1. Backup links [1](https://1drv.ms/u/s!Aok6lAYtb5vYzFpoUwuhR24el4xr?e=deWaV1), [2](https://kuleuven-my.sharepoint.com/:u:/g/personal/hui_zhang_kuleuven_be/EWqD4-A8Hy5IrFRo-6aKQN4BX6hK5GQ_6gOiBRgY0WCVmQ?e=oDFLn0) to download our mini dataset.

# Citation

@ARTICLE{Hui_GA_CNN,  
  author    = {Zhang, Hui and Peeters, Jef and Demeester, Eric and Kellens, Karel},  
  journal   = {IEEE Transactions on Robotics},   
  title     = {Deep Learning Reactive Robotic Grasping With a Versatile Vacuum Gripper},   
  year      = {2023},  
  volume    = {39},  
  number    = {2},  
  pages     = {1244-1259},  
  doi       = {10.1109/TRO.2022.3226148}  
 }