yolov8-训练自己的数据集

---

下载网址

yolov8

https://github.com/ultralytics/ultralytics

---

环境配置

Python>=3.8

在线安装：PyTorch>=1.8 — 离线安装PyTorch>=1.8

pip install -r requirements.txt
pip install yolo
pip install ultralytics

---

数据集分配

images--图片
labels--txt
此处本人依据coco数据集的格式，下面有划分代码

---

创建 yaml 文件

此处本人依据coco数据集的格式，相对路径和绝对路径都可

---

检测自己环境是否搭建完成

yolo predict model=yolov8s.pt source=ultralytics/assets/bus.jpg

没有权重文件不用担心，会自动下载

---

训练自己的数据集

yolo detect train data=datasets/superbus/coco128.yaml model=yolov8s.pt epochs=300 batch=4 lr0=0.01
yolo detect train data=datasets/superbus/coco128.yaml model=yolov8s.pt epochs=300 batch=4 workers=0 device=0

---

yolo 指令

	yolo 任务 模式 参数
	
	其中	任务（可选）是[detect, segment, classify]中的一个
		    模式（必需）是[train, val, predict, export, track]中的一个
		    参数（可选）是任意数量的自定义“arg=value”对，如“imgsz=320”，可覆盖默认值。

---

代码 xml --> txt

# -*- coding: utf-8 -*-
import xml.etree.ElementTree as ET
import os
from os import getcwd

sets = ['train', 'val', 'test']
classes = ['warn', 'instructions', 'limit', 'forbid']
# 改成自己的类别
abs_path = os.getcwd()
print(abs_path)


def convert(size, box):
    dw = 1. / (size[0])
    dh = 1. / (size[1])
    x = (box[0] + box[1]) / 2.0 - 1
    y = (box[2] + box[3]) / 2.0 - 1
    w = box[1] - box[0]
    h = box[3] - box[2]
    x = x * dw
    w = w * dw
    y = y * dh
    h = h * dh
    return x, y, w, h


# 我比较懒，下面几乎用的都是绝对路径
def convert_annotation(image_id):
    in_file = open('D:/project/project/yolov5-Road-detection/data/Annotations/%s.xml' % (image_id),
                   encoding='UTF-8')
    out_file = open('D:/project/project/yolov5-Road-detection/data/labels/%s.txt' % (image_id), 'w')
    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
        # difficult = obj.find('Difficult').text
        cls = obj.find('name').text
        if cls not in classes or int(difficult) == 1:
            continue
        cls_id = classes.index(cls)
        xmlbox = obj.find('bndbox')
        b = (float(xmlbox.find('xmin').text), float(xmlbox.find('xmax').text), float(xmlbox.find('ymin').text),
             float(xmlbox.find('ymax').text))
        b1, b2, b3, b4 = b
        # 标注越界修正
        if b2 > w:
            b2 = w
        if b4 > h:
            b4 = h
        b = (b1, b2, b3, b4)
        bb = convert((w, h), b)
        out_file.write(str(cls_id) + " " + " ".join([str(a) for a in bb]) + '\n')


wd = getcwd()
for image_set in sets:
    if not os.path.exists('D:/project/project/yolov5-Road-detection/data/labels/'):
        os.makedirs('D:/project/project/yolov5-Road-detection/data/labels/')
    image_ids = open('D:/project/project/yolov5-Road-detection/data/ImageSets/Main/%s.txt' % (
        image_set)).read().strip().split()

    if not os.path.exists('D:/project/project/yolov5-Road-detection/data/dataSet_path/'):
        os.makedirs('D:/project/project/yolov5-Road-detection/data/dataSet_path/')

    list_file = open('dataSet_path/%s.txt' % (image_set), 'w')
    # 这行路径不需更改，这是相对路径
    for image_id in image_ids:
        list_file.write('D:/project/project/yolov5-Road-detection/data/images/%s.jpg\n' % (image_id))
        convert_annotation(image_id)
    list_file.close()

---

代码 split_train_val

# coding:utf-8

import os
import random
import argparse
import shutil

random.seed(666)
parser = argparse.ArgumentParser()
# image，xml--数据集的路径
parser.add_argument('--xml_path', default='G:/车辆和车牌数据集v2.0/images', type=str, help='input xml label path')
# 保存的地址，选择自己数据下的ImageSets/Main--
parser.add_argument('--save_path', default='G:/车辆和车牌数据集v2.0/ImageSets/Main_images', type=str,
                    help='output txt label path')
opt = parser.parse_args()

trainval_percent = 0.8  # 训练集和验证集所占比例

xml_file_path = opt.xml_path
save_path = opt.save_path
save_path_train = save_path + '/train2017'
save_path_val = save_path + '/val2017'

total_xml = os.listdir(xml_file_path)
os.makedirs(save_path, exist_ok=True)
os.makedirs(save_path_train, exist_ok=True)
os.makedirs(save_path_val, exist_ok=True)

num = len(total_xml)
list_index = range(num)
tv = int(num * trainval_percent)

train_8 = random.sample(list_index, tv)

for i in list_index:
    if i in train_8:
        shutil.copy(xml_file_path + '/' + total_xml[i], save_path_train)
    else:
        shutil.copy(xml_file_path + '/' + total_xml[i], save_path_val)

• 参考文献：https://docs.ultralytics.com/zh/