【EAST】环境变量配置！！

1.linux下EAST环境变量配置

1. 库的安装首选conda，再考虑pip，https://pypi.org/

2. conda create -n ghEAST python==3.6

   

3. 进入虚拟环境

   

4. opencv安装

   pip3 install opencv-python==4.0.0.21

   

5. pytorch安装

   conda install pytorch==1.0.1

   

6. shapely

   conda install Shapely==1.6.4

7. lanms

   pip3 install lanms==1.0.2

8. 安装PIL时如果使用conda install PIL 会出现

   [外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传(img-D2nJpkBI-1606473612929)(C:\Users\pc\AppData\Roaming\Typora\typora-user-images\image-20201127102221329.png)]

   

   

pil只支持python2.7

用pillow代替PIL

9. pytorch1.0会出现no model named “torch.utils.tensorboard"

   

   解决办法：升级pytorch版本 conda install pytorch=1.1

   所以开始的时候直接conda install pytorch=1.1，节省时间！！！

10. 之后运行train_inner.py会报错

    Traceback (most recent call last):
    File “train_inner.py”, line 2, in
    import torch
    File “/home/guohui/anaconda3/envs/ghEAST/lib/python3.6/site-packages/torch/init.py”, line 79, in
    from torch._C import *

    cuda的库文件读取不了了

    解决办法：conda install cudatoolkit==9.0

    

    cuda版本要对应

    查看对应版本办法 cat /usr/local/cuda/version.txt

    

11. no model named “past”

    

    解决：pip3 install future

    

12. 

    tensorflow与numpy版本不兼容

    

13.

解决办法：图片数量能被batch-size整除，改变图片数量或者batch-size大小