visual-pushing-grasping环境配置及复现

visual-pushing-grasping环境配置及复现 - 知乎

视觉机械臂 visual-pushing-grasping - 云+社区 - 腾讯云

Requirements: (tested on Ubuntu 16.04.4 LTS):

Python 2.7 or Python 3

    NumPy, SciPy, OpenCV-Python, Matplotlib. You can quickly install/update these dependencies by running the following (replace pip with pip3 for Python 3):

    pip install numpy scipy opencv-python matplotlib

    PyTorch version 0.3. Since 0.3 is no longer the latest version, see installation instructions here or run the following:

    pip install torch==0.3.1 torchvision==0.2.0

    Support for PyTorch version 0.4+ is work-in-progress and lives in this branch, but currently remains unstable.

    V-REP (simulation environment)

配置开始

环境： conda create -n vpg python=3.6
           //source activate(importent)
           source activate vpg
    安装依赖：pip install numpy scipy opencv-python matplotlib     
安装pytorch： conda install pytorch==1.4.0 torchvision==0.5.0 cudatoolkit=10.1 -c pytorch
    安装add/coppeliaSim: https://www.icode9.com/content-3-801529.html
          下载官方V-REP，参考网址：https://www.coppeliarobotics.com/previousVersions;
          选择中间EDU，Ubuntu16.04版本，是免费的；
          将安装包放到一定的路径下，这里放到"/Home/heixiazi/add"目录下，并且重命名为coppeliaSim;
          将下载好的安装包解压到/home/heixiazi/add/CoppeliaSim目录下；
          添加source源设置coppeliaSim路径: sudo gedit ~/.bashrc
          在.bashrc中最后一行添加:export VREP_ROOT=$HOME/heixiazi/add/CoppeliaSim  
          source ~/.bashrc
          启动V-REP: cd到安装目录下
                     启动sh文件 ./coppeliaSim.sh

运行

1.  Checkout this repository and download our pre-trained models.
    git clone https://github.com/andyzeng/visual-pushing-grasping.git visual-pushing-grasping
    cd visual-pushing-grasping/downloads
    ./download-weights.sh
    cd ..
2.  Run V-REP (navigate to your V-REP directory and r