1、环境安装
在github上下载ultralytics框架源码，安装库参照requestment.txt，yolov8对python版本有一定要求。如果使用pip install ultralytics==8.0.0，那么这个库会存放在环境/site_packages里面，更改源码时候，到该路径下更改。

2、数据处理
数据处理的方式和其它版本yolo一样，生成标签的txt文件，在ultralytics/cfg/datasets里面，设定自己的数据集yaml文件。
data.yaml

第一种写法
path: D:\dataset\Test  # dataset root dir
train: images/train  # train images (relative to 'path')
val: images/val  # val images (relative to 'path')
test:  # test images (optional)
# Classes
names:
  0: C1
  1: C2

第二种写法
train: images/train.txt  # train images (relative to 'path')
val: images/val.txt  # val images (relative to 'path')
test: images/test.txt# test images (optional)
nc:1   #num of class
# Classes
names:["c1","c2"]

3、模型训练
训练前：
1、配置ultralytics/models/v8/yolov8.yaml中的类别数目
2、配置ultralytics/cfg/default.yaml中的任务，以及其它配置信息
官方训练、验证、测试、模型导出代码：

yolo task=detect    mode=train    model=yolov8n.pt        args...
          classify       predict        yolov8n-cls.yaml  args...
          segment        val            yolov8n-seg.yaml  args...
                         export         yolov8n.pt        format=onnx  args...
              
              
训练：yolo task=detect mode=train model=weights/yolov8n.pt data=data/mydata.yaml batch=16 epochs=50 imgsz=640 workers=16 device=0
验证：yolo task=detect mode=val model=runs/detect/train3/weights/best.pt data=data/mydata.yaml device=0
预测：yolo task=detect mode=predict model=weights/best.pt source=data/test_images save=True

第二种写法：

# train model
from ultralytics import YOLO
# Load a model 三选一
model = YOLO('yolov8n.yaml')  # build a new model from YAML 配置好模型和训练数据yaml文件信息
model = YOLO('yolov8n.pt')  # load a pretrained model (recommended for training)
model = YOLO('yolov8n.yaml').load('yolov8n.pt')  # build from YAML and transfer weights
# Train the model
model.train(data='coco128.yaml', epochs=100, imgsz=640)

#val model
from ultralytics import YOLO
# Load a model  二选一
model = YOLO('yolov8n.pt')  # load an official model
model = YOLO('path/to/best.pt')  # load a custom model
# Validate the model
metrics = model.val()  # no arguments needed, dataset and settings remembered
metrics.box.map    # map50-95
metrics.box.map50  # map50
metrics.box.map75  # map75
metrics.box.maps   # a list contains map50-95 of each category

#predict 
from ultralytics import YOLO
import cv2
import os,datetime
os.environ['KMP_DUPLICATE_LIB_OK']='True'    #如果训练时候报Initializing libiomp5md.dll, but found libiomp5md.dll already initialized错误，加上这句
model = YOLO('best.pt') #已经训练好的模型
# Define path to the image file
source = 'img or video' #待预测的数据保存路径
# Run inference on the source
results = model(source, mode='predict', save=True)  # list of Results objects

#export model
from ultralytics import YOLO
# Load a model
model = YOLO('yolov8n.pt')  # load an official model
model = YOLO('path/to/best.pt')  # load a custom trained
# Export the model
model.export(format='onnx')

自己做的检测代码，实现boxes的显示，可根据需求进行修改：

from ultralytics import YOLO
import cv2
import os,datetime,time
os.environ['KMP_DUPLICATE_LIB_OK']='True'
path_pt="x.pt"
model=YOLO(path_pt)
img_dir="./img"
image_files=[os.path.join(img_dir,file) for file in os.listdir(img_dir) if file.endswith((".jpg","png","jpeg","mp4"))]
for img_file in image_files:
	mat_img=cv2.imread(img_file)
	res=model(img_file,save=True,conf=0.4,iou=0.2,imgsz=640,device="cuda:0")  #use cuda
	res_coord=[]
	outputs=res[0]
	boxes=res[0].boxes  # include xyxy xywh xywhn
	boxes=boxes.xyxy
	boxes=boxes.cpu()  #cuda data transfer cpu data
	boxes=boxes.tolist()
	count=0
	for box in boxes:
		count+=1
		width=int(box[2])-int(box[0])
		height=int(box[3])-int(box[1])
		x_center=(int(box[2])+int(box[0]))//2
		y_center=(int(box[3])+int(box[1]))//2
		res_coord.append((x_center,y_center))
		img=cv2.circle(mat_img,(x_center,y_center),5,(0,0,255),-1)
	date_time=datetime.datetime.strftime(datetime.datetime.now(),"%Y%m%d%H%M%S")
	path=".res/"+date_time+".jpg"
	cv2.imwrite(path,img)
	res.coord.append(path)  # include box(center) and the result path of img