bairenchen / ee-cs-rookie Goto Github PK

Reference:
https://blog.csdn.net/virus1175/article/details/104577574
https://blog.csdn.net/weixin_43325818/article/details/107123611

pip install torch==1.2.0 torchvision==0.4.0
ref: https://pytorch.org/get-started/previous-versions/

Problem descriptions:
Loaded runtime CuDNN library: 7.1.4 but source was compiled with: 7.6.5. CuDNN library major and minor version needs to match or have higher minor version in case of CuDNN 7.0 or later version. If using a binary install, upgrade your CuDNN library. If building from sources, make sure the library loaded at runtime is compatible with the version specified during compile configuration.

ref:
[important!]https://askubuntu.com/questions/767269/how-can-i-install-cudnn-on-ubuntu-16-04
https://blog.csdn.net/wanzhen4330/article/details/81704474
https://blog.csdn.net/zong596568821xp/article/details/86098833
tensorflow/tensorflow#35376
https://developer.nvidia.com/rdp/cudnn-archive
https://jingyan.baidu.com/article/d8072ac40dd3b0ac95cefdf7.html
https://stackoverflow.com/questions/61039316/loaded-runtime-cudnn-library-7-1-2-but-source-was-compiled-with-7-6-0-ubuntu

The console window in Pycharm is closed by default, unlike Spyder.
The solution is to click Run in the menu bar, choose Edit Configurations, tick Run with Python Console, and apply.

Reference: https://blog.csdn.net/qq_38355352/article/details/88808522

https://blog.csdn.net/qq_17783559/article/details/87932760

1 -> https://keras.io/api/metrics/classification_metrics/
2 -> https://pypi.org/project/keras-metrics/

（1）Download packages from https://share.weiyun.com/5Hm7kxy.
ref: https://blog.csdn.net/qq_43060552/article/details/103189040
（2）Copy it to the corresponding category.
e.g. scp -r /home/cbr/cnn_graph/tutorials /usr/local/lib/python3.6/dist-packages/tensorflow_core/examples

https://www.jianshu.com/p/6b998ba2c6a6

I strongly recommend that python environment in your PC is based on Anaconda3.

1. Install Anaconda3 & Pycharm.
2. Open Pycharm, create new project, click Existing interpreter, and choose the location of your anaconda python. (i.e. D:\Applications\Anaconda3\python.exe)
3. Reboot Pycharm.

If you have created your own projects based on other python environments, the solution steps: File->Settings->Project: XXX->Project Interpreter->the location of anaconda python.

Tips:

1. In Prompt, you can also use pip to install packages and the packages are still in conda list.
2. The solution of changing background color: File->Settings->Editor->Color Scheme->General->Darcula.

I have installed and implemented Anaconda3 and PyCharm2019 in my PC. The environment configuration of PyCharm is based on the conda list in Anaconda3.
In order to learn image process, I WANTED TO INSTALL OPENCV IN CONDA LIST.
Using
conda install -chttps://conda.binstar.org/menpo opencv
conda install -c menpo opencv
conda install -c conda-forge opencv
conda install -c conda-forge/label/broken opencv
didn't work.

Here's the solution:

1. Download opencv .whl File

According to https://blog.csdn.net/iracer/article/details/90746292, I downloaded the .whl file (opencv_python-4.2.0.34-cp37-cp37m-win_amd64.whl) from https://pypi.tuna.tsinghua.edu.cn/simple/opencv-python/.
(Here's an another address of downloading opencv but the speed was too slow, https://blog.csdn.net/iracer/article/details/90746292)

2. Unzip the .whl File

I put the .whl file into Anaconda3 third-party package directory, ...\Anaconda3\Lib\site-packages.
(My path: D:\Applications\Anaconda3\Lib\site-packages)

Ran cmd, and input cd /d D:\Applications\Anaconda3\Lib\site-packages
Note: /d is important!!!
(Reference of cd to the specified directory: https://blog.csdn.net/relijin/article/details/62416267 or https://blog.csdn.net/qq_32160853/article/details/81303127?utm_medium=distribute.pc_relevant_t0.none-task-blog-BlogCommendFromMachineLearnPai2-1.nonecase&depth_1-utm_source=distribute.pc_relevant_t0.none-task-blog-BlogCommendFromMachineLearnPai2-1.nonecase)

Input pip install opencv_python-4.2.0.34-cp37-cp37m-win_amd64.whl

It worked!!!

3. Check

Checked the environment configuration:

• Ran Anaconda Prompt, input conda list and checked whether opencv-python is in conda list;
• Ran PyCharm, File-Settings-Project-Project Intepreter, and searched opencv-python package, normally the file should exist.

==================== update at 7/1/2020 ====================

In SIFT algorithm, cv2.xfeatures2d.SIFT_create() was excluded and couldn't run (I searched and knew that this algorithm was patented).

By https://stackoverflow.com/questions/52305578/sift-cv2-xfeatures2d-sift-create-not-working-even-though-have-contrib-instal, I knew that opencv version <=3.4.2.16 still contained SIFT algorithm.

At the same time, I noticed little difference --> contrib. I searched and got the answer at https://blog.csdn.net/qq_43258953/article/details/104760310

The solution

1. Uninstall the Existing Version of Opencv

Ran cmd, input cd /d D:\Applications\Anaconda3\Lib\site-packages, input pip uninstall opencv_python,
checked whether the package was moved from site-packages. (https://blog.csdn.net/roytao2/article/details/53434096)

2. Download the Low Version and Unzip it at the Specific Directory

Download it from ​​https://pypi.org/project/opencv-contrib-python/3.4.2.16/#files (opencv_python https://pypi.org/project/opencv-python/3.4.2.16/#files), put it in my path: D:\Applications\Anaconda3\Lib\site-packages. (I only downloaded contrib verison.)

Input pip install opencv_contrib_python-3.4.2.16-cp37-cp37m-win_amd64.whl, checked it in the end.

Finished!!!

cd ...
rm ... (e.g. insu_dset.npz)

https://blog.csdn.net/HighDS/article/details/104204632

https://blog.csdn.net/qq_31456593/article/details/89396475

https://blog.csdn.net/SherryHuang94/article/details/79629206

https://blog.csdn.net/whatday/article/details/53696288

https://blog.csdn.net/HighDS/article/details/104204632

solution

https://blog.csdn.net/SherryHuang94/article/details/79629206

https://blog.csdn.net/xpy870663266/article/details/103235073

https://blog.csdn.net/yeler082/article/details/83818771
https://blog.csdn.net/baidu_39622935/article/details/84299439

ref : https://blog.csdn.net/u012114438/article/details/104425573

how to check out the version information for the Ubuntu system?
ref: https://jingyan.baidu.com/article/ca41422f0ab38f1eae99edc3.html

1. Establish soft link between cuda-10.1 and cuda.
2. Establish soft link as below.
   tensorflow/tensorflow#39132
   sudo ln -s /usr/local/cuda-10.1/targets/x86_64-linux/lib/libcudart.so.10.1 /usr/lib/x86_64-linux-gnu/libcudart.so.10.1
   All the problems were solved.
3. Ehoose GPU 0 or 1 to run data.
   https://stackoverflow.com/questions/64956013/how-can-i-select-a-specific-gpu-with-tensorflow-and-yolov3-tensorflow-gpu-2-3
   tf.config.set_visible_devices(physical_devices[0:1], 'GPU')

https://blog.csdn.net/qq_15971883/article/details/97185112

from keras import ... -> from tensorflow.keras import ...
ref: https://www.jianshu.com/p/9da54361d289

(1) sudo su -> password -> root
(2) vi ~/.bashrc -> add below
export PATH=/usr/local/cuda-10.0/bin${PATH:+:${PATH}}
export LD_LIBRARY_PATH=/usr/local/cuda-10.0/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}
export CUDA_HOME=/usr/local/cuda-10.0
(3) cd /usr/local -> sudo rm -rf cuda -> sudo ln -s /usr/local/cuda-10.0 /usr/local/cuda
(4) source ~/.bashrc
(5) check nvcc -V
ref: https://blog.csdn.net/c2250645962/article/details/105177052/

(1) git clone
(2) VCS -> Git -> Clone
(3) File -> Setting -> Projects: xxx -> Project Interpreter
(4) Tools -> Deployment -> Configuration -> Mappings
(4) Ctrl+Shift+Alt+x