用yolo8训练自己的模型

YOLOv8训练自己的数据集+常用传参说明-阿里云开发者社区 (aliyun.com)这个链接是教程

1.在其中链接中 转换数据集格式用完这个之后用以下代码，这个代码是转换格式将xml转为txt

格格巫:

# OBB xml(rolabelImg)->txt(yolo)
# -*- coding: utf-8 -*-
import xml.etree.ElementTree as ET
import os
import math
from os import getcwd

classes1_5 = ['bear','wineglass','clock']  # 改成自己的类别

def convert(size, box):
    dw = 1. / (size[0])
    dh = 1. / (size[1])
    x = box[0] * dw
    y = box[1] * dh
    w = box[2] * dw
    h = box[3] * dh
    return x, y, w, h


def convert_annotation(image_id):
    in_file = open(r'D:\jiqiren\envs\yolov8\Lib\site-packages\ultralytics\data\Annotations\%s.xml' % (image_id), encoding='UTF-8')    # 存放xml标签的文件夹
    out_file = open(r'D:\jiqiren\envs\yolov8\Lib\site-packages\ultralytics\data\labels\%s.txt' % (image_id), 'w')    # 转换出的txt标签存放文件夹
    tree = ET.parse(in_file)
    root = tree.getroot()
    size = root.find('size')
    w = int(size.find('width').text)
    h = int(size.find('height').text)
    for obj in root.iter('object'):
        difficult = obj.find('difficult').text
        cls = obj.find('name').text
        if cls not in classes1_5 or int(difficult) == 1:
            continue
        cls_id = classes1_5.index(cls)
        xmlbox = obj.find('robndbox')
        b = (float(xmlbox.find('cx').text), float(xmlbox.find('cy').text), float(xmlbox.find('w').text),
             float(xmlbox.find('h').text))
        theta = float(xmlbox.find('angle').text)
        b1, b2, b3, b4 = b
        if b3 < b4:
            b = (b1, b2, b4, b3)
            theta = int(((theta * 180 / math.pi) + 90) % 180)
        else:  # 如果有b3≈b4, 那么OBB是正方形, 两个theta值都适用, 但是不利于训练
            theta = int(theta * 180 / math.pi)
        bb = convert((w, h), b)
        # out_file.write(str(cls_id) + " " + " ".join([str(a) for a in bb]) + " " + str(theta) + '\n')
        out_file.write(str(cls_id) + " " + " ".join([str(a) for a in bb]) + '\n')

if not os.path.exists('labels/'):
    os.makedirs('labels/')    # yolo格式的标签存放地址
fileDir = r"D:\jiqiren\envs\yolov8\Lib\site-packages\ultralytics\data\Annotations"
image_name = os.listdir(fileDir)    # 获取指定文件夹下所有文件名，含后缀
image_ids = [x.split('.')[0] for x in image_name] 
print(image_ids)
for image_id in image_ids:
    convert_annotation(image_id)
print("转换完成")

如果有问题使用以下代码

格格巫:

# OBB xml(rolabelImg)->txt(yolo)
# -*- coding: utf-8 -*-
import xml.etree.ElementTree as ET
import os
import math
from os import getcwd

classes1_5 = ['bear', 'wineglass', 'clock']  # 改成自己的类别

def convert(size, box):
    dw = 1. / (size[0])
    dh = 1. / (size[1])
    x = box[0] * dw
    y = box[1] * dh
    w = box[2] * dw
    h = box[3] * dh
    return x, y, w, h

def convert_annotation(image_id):
    try:
        in_file_path = r'D:\jiqiren\envs\yolov8\Lib\site-packages\ultralytics\data\Annotations\%s.xml' % (image_id)
        out_file_path = r'D:\jiqiren\envs\yolov8\Lib\site-packages\ultralytics\data\labels\%s.txt' % (image_id)
        
        if not os.path.exists(in_file_path):
            print(f"文件未找到: {in_file_path}")
            return
        
        with open(in_file_path, encoding='UTF-8') as in_file:
            tree = ET.parse(in_file)
            root = tree.getroot()
            size = root.find('size')
            w = int(size.find('width').text)
            h = int(size.find('height').text)
            
            with open(out_file_path, 'w') as out_file:
                for obj in root.iter('object'):
                    difficult = obj.find('difficult').text
                    cls = obj.find('name').text
                    if cls not in classes1_5 or int(difficult) == 1:
                        continue
                    cls_id = classes1_5.index(cls)
                    xmlbox = obj.find('robndbox')
                    b = (float(xmlbox.find('cx').text), float(xmlbox.find('cy').text), float(xmlbox.find('w').text),
                         float(xmlbox.find('h').text))
                    theta = float(xmlbox.find('angle').text)
                    b1, b2, b3, b4 = b
                    if b3 < b4:
                        b = (b1, b2, b4, b3)
                        theta = int(((theta * 180 / math.pi) + 90) % 180)
                else:  # 如果有b3≈b4, 那么OBB是正方形, 两个theta值都适用, 但是不利于训练
                        theta = int(theta * 180 / math.pi)
                    bb = convert((w, h), b)
                    # out_file.write(str(cls_id) + " " + " ".join([str(a) for a in bb]) + " " + str(theta) + '\n')
                    out_file.write(str(cls_id) + " " + " ".join([str(a) for a in bb]) + '\n')
    except Exception as e:
        print(f"转换 image ID {image_id} 的注释时出错: {e}")

if not os.path.exists(r'D:\jiqiren\envs\yolov8\Lib\site-packages\ultralytics\data\labels'):
    os.makedirs(r'D:\jiqiren\envs\yolov8\Lib\site-packages\ultralytics\data\labels')  # yolo格式的标签存放地址

fileDir = r"D:\jiqiren\envs\yolov8\Lib\site-packages\ultralytics\data\Annotations"
image_name = os.listdir(fileDir)  # 获取指定文件夹下所有文件名，含后缀
image_ids = [x.split('.')[0] for x in image_name]  # 去除文件名列表的后缀
print(image_ids)

for image_id in image_ids:
    convert_annotation(image_id)
print("转换完成")

注意：链接中的voc_label.py文件转换出的txt可能是空文档，以及其中第58改为

 list_file.write(abs_path + '\data\images\%s.jpg\n' % (image_id))

2.训练完之后进行测试

上传进行图片测试

文件运行
import cv2
from ultralytics import YOLO



if __name__ == '__main__':

    pth_path = "D:\jiqiren\envs\yolov8\Lib\site-packages\\ultralytics\\runs\detect\\train18\weights\\best.pt"

    test_path =r"C:\Users\LENOVO\OneDrive\桌面\xie\47549eefac88b00bb5731496088f7dd.jpg"
    # Load a model
    # model = YOLO('yolov8n.pt')  # load an official model
    model = YOLO(pth_path)  # load a custom model

    # Predict with the model
    results = model(test_path)

    for result in results:
        boxes = result.boxes  # Boxes object for bounding box outputs
        masks = result.masks  # Masks object for segmentation masks outputs
        keypoints = result.keypoints  # Keypoints object for pose outputs
        probs = result.probs  # Probs object for classification outputs
        obb = result.obb  # Oriented boxes object for OBB outputs
        result.show()  # display to screen# predict on an image

打开摄像头进行测试

import cv2
from ultralytics import YOLO

# 加载预训练的YOLOv8模型
model = YOLO('D:\jiqiren\envs\yolov8\Lib\site-packages\\ultralytics\\runs\detect\\train22\weights\\best.pt')  # 你可以选择不同的预训练模型，如'yolov8s.pt', 'yolov8m.pt'等

# 打开摄像头
cap = cv2.VideoCapture(0)  # 0代表默认摄像头，如果有多个摄像头，可以尝试1, 2等

while True:
    # 读取摄像头帧
    ret, frame = cap.read()
    if not ret:
        break

    # 使用YOLOv8进行检测
    results = model(frame)

    # 绘制检测结果
    annotated_frame = results[0].plot()

    # 显示结果
    cv2.imshow('YOLOv8 Real-time Detection', annotated_frame)

    # 按'q'键退出循环
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

# 释放摄像头并关闭窗口
cap.release()
cv2.destroyAllWindows()

3.使用labelimg进行标注，标注不能旋转，可以找rolabelimg进行标注

"C:\Users\dell\AppData\Local\Temp\360zip$Temp\360$1\roLabelImg-master"

打开roLabelImg-master文件夹，打开cmd,打开你配置的环境conda activate -环境名字-,输入python roLabelImg.py，就可以使用啦