[yolov5] environment construction: win11 + mx450

1. Computer configuration

2. install cuda and cndua

2.1 CUDA

CUDA（Compute Unified Device Architecture）, It's the video card manufacturer NVIDIA The new computing platform . according to GPU Of driver The version download corresponds to CUDA, Can be in NVIDIA Control panel - help - system information - Component driver edition , My version is 11.4

Support 11.4 Of cuda, Can be in https://developer.nvidia.com/cuda-toolkit-archive Download

Be careful ,win10 Version of CUDA And win11 compatible ,PyTorch And CUDA There is a corresponding relationship between , The high version of the CUDA Corresponding to the higher version PyTorch, But higher than 1.10 Version of torch In the training YOLO5 When it comes to P=0, R=0 Of bug, Because it was just installed 10.4 Version of CUDA, Later, it was reinstalled

Direct installation 10.2 Version of CUDA：https://developer.nvidia.com/cuda-10.2-download-archive

Download and install , Can be in Set up - System - system information - Advanced system setup - Two new items are added in the environment variable

---

2.2 cuDNN

cuDNN It can speed up the training of the model , Optimize GPU performance , Download directly with CUDA Corresponding version cuDNN that will do ：

Unzip after downloading , Put the extracted three folders into CUDA Installation directory C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.2

such ,CUDA and cuDNN Just install it ！

3. download YOLO v5 Source code

stay Github Download the official YOLO v5 Code , If you cannot access it, you can use the image source ,click me

You can look at them carefully README.md file , There are instructions about the environment and operation methods , such as Python Version and dependent library installation ：

Clone repo and install requirements.txt in a Python>=3.7.0 environment, including PyTorch>=1.7.

git clone https://github.com/ultralytics/yolov5  # clone
cd yolov5
pip install -r requirements.txt  # install

4. Conda Installation and environment creation of

4.1 install Miniconda

conda Can be used to create virtual environments , If you need to run different models , These models have different configuration requirements for the environment , In particular, different models may require different versions of dependent libraries , Reloading the environment is troublesome , Use at this time conda Switch environment like a duck to water ！

install Miniconda, Because the local installation is Python3.7, So I downloaded 3.7 Version of （ In this case , Use conda The default virtual environment is Python3.7, You can download other versions according to your needs ）

It should be noted that , You need to add environment variables during installation ：

Check... After installation environment variable , Refer to the third figure , After installation, enter it on the command line conda, If there is output and no error is reported, the installation is successful ！

---

4.2 install Python Environmental Science

Next use conda Command to create virtual environment , For ease of operation , Let's first enter the decompressed yolov5 Under the table of contents , Then use the following command ：

conda create -n yolov5 python=3.7   #  establish  Python3.7 Of  yolov5  A virtual environment 
conda activate yolov5               #  Activate  yolov5  Environmental Science 

At this time, we are in yolov5 Environment , Because at this time, it is in the decompressed yolov5 Under the table of contents , You can directly access the files in the directory , According to the official website README Install the dependent libraries according to the prompts of the dependent libraries ：

First, according to what you downloaded CUDA, Install the appropriate version of pytorch（CUDA10.2 Download the following version pytorch That's all right. , Other versions of CUDA You can download it on the official website pytorch）：

pip install torch==1.9.1+cu102 torchvision==0.10.1+cu102 torchaudio==0.9.1 -f https://download.pytorch.org/whl/torch_stable.html

next , We turn on yolov5 In the catalog requirements.txt file , Comment out the following two libraries （ Because it has just been installed ）：

#torch>=1.7.0
#torchvision>=0.8.1

Then use the following command to install the remaining libraries ：

pip install -r requirements.txt  # install

thus , The environment is all set up , Let's run it ！

---

At present yolov5 In the environment , Enter the command

python detect.py --source 0

This command uses the trained model to detect the object captured by the camera , Normal operation indicates that the environment is ok , Anyway, mine is ok

---

4.3 conda command

To this step ,conda You should learn how to use , After creating the environment , Use activate Command to activate an environment , Switch to a specific directory , You can run code with specific environmental requirements .

conda More common commands of are as follows ( Don't take notes when you enter an order )：

conda create --name xxx --clone xxx	# clone
conda remove --name <OldName> --all	# remove
conda create -n yolov5 python=3.7   #  establish  Python3.7 Of  yolov5  A virtual environment 
conda activate yolov5               #  Activate  yolov5  Environmental Science 
conda activate xxxx               	# Turn on xxxx Environmental Science 
conda deactivate                  	# Shut down the environment 
conda env list                    	# Show all virtual environments 
conda info --envs                 	# Show all virtual environments 

---

REFERENCE
1 . conda command
2 . yolov5
3 . solve YOLOV5 Appear all for nan and 0 The problem of

Next ： be based on YOLOV5 Of SAR Image ship detection