引言

笔者YOLOv5专栏链接导航：

1. 【YOLOv5】6.0环境搭建（不定时更新）
2. 【YOLOv5】训练自己的数据集

The author has always been very partialyolov5系列,Because it is very friendly,上手容易,Compatibility is also relatively high（There won't be many demanding requirements,Especially for training hardware configuration）,From building the environment to training to application,都非常简便.Let users spend more time on coordination and improvement of actual application scenarios.
yolov5的最新版本6.1
源码：https://github.com/ultralytics/yolov5/tree/v6.1

Some updated information notes：https://github.com/ultralytics/yolov5/releases/tag/v6.1
Author update date : 2022.3.2

1.环境搭建

conda create -n yolov5-6 python=3.7
conda activate yolov5-6

官网的安装教程：

Generally run to downloadtorch会非常慢,So it is best to install it separately firsttorch

The author's installation instruction flow is as follows：

git clone https://github.com/ultralytics/yolov5  # 1
cd yolov5  # 2
pip install -r requirements.txt  # 3

Follow the prompts to decide which version to install,如上图所示
官网安装命令 https://pytorch.org/get-started/locally/

# 单独安装pythorch(gpu版本)
install torch==1.10.2+cu102 torchvision==0.11.3+cu102 torchaudio===0.10.2+cu102 -f https://download.pytorch.org/whl/cu102/torch_stable.html  # 4
# Run the official installation instructions again,At this point the installation will be much faster
pip install -r requirements.txt  # 5

If other necessary libraries are encountered, the download is very slow,Then you can install and use the mirror source separately

pip install xxx -i https://pypi.tuna.tsinghua.edu.cn/simple
pip install xxx -i https://pypi.douban.com/simple
# 然后再
pip install -r requirements.txt

requirements.txt的内容如下：

2.下载预训练模型

链接（Scroll to the bottom of the page）：预训练模型

3. 推理

任意新建一个.py文件,写入如下的代码,在终端运行

import torch

# Model
model = torch.hub.load('ultralytics/yolov5', 'yolov5s')  # or yolov5m, yolov5x, custom
# Images
# img = 'https://ultralytics.com/images/zidane.jpg' # or file, PIL, OpenCV, numpy, multiple
img = 'data/images/zidane.jpg'
# Inference
results = model(img)
# Results
results.print()  # or .show(), .save(), .crop(), .pandas(), etc.

显示结果为：

4. 测试

--project指定保存路径;--weights指定测试模型

python detect.py --source data/images/zidane.jpg --weights v5_pre_models/yolov5s.pt --project out --img-size 640