Yolov5训练自己的数据集（超详细）

一、从官网下载最新的yolov5代码
二、新建VOCData文件夹

三、VOCData文件夹结构

1、新建Annotations文件夹，存放标签

简单的xml文件，应该长这样

复杂的xml文件，应该长这个样子

2、新建images文件夹，存放图片数据

注意：需要观察自己的图片文件的后缀名，后面需要用到，不然可能出现程序找不到图片的情况
3、新建split_train_val.py，用来划分训练集与验证集

#coding:utf-8

import os
import random
import argparse

parser = argparse.ArgumentParser()#xml文件的地址，根据自己的数据进行修改 xml一般存放在Annotations下
parser.add_argument('--xml_path',default='Annotations', type=str, help='input xml label path')
#数据集的划分，地址选择自己数据下的ImageSets/Main
parser.add_argument('--txt_path',default='ImageSets/Main', type=str, help='output txt label path')
opt = parser.parse_args()

trainval_percent =0.7  # 训练集和验证集所占比例。 这里没有划分测试集
train_percent =0.5     # 训练集所占比例，可自己进行调整
xmlfilepath = opt.xml_path
txtsavepath = opt.txt_path
total_xml = os.listdir(xmlfilepath)if not os.path.exists(txtsavepath):
    os.makedirs(txtsavepath)

num =len(total_xml)
list_index =range(num)
tv =int(num * trainval_percent)
tr =int(tv * train_percent)
trainval = random.sample(list_index, tv)
train = random.sample(trainval, tr)

file_trainval =open(txtsavepath +'/trainval.txt','w')
file_test =open(txtsavepath +'/test.txt','w')
file_train =open(txtsavepath +'/train.txt','w')
file_val =open(txtsavepath +'/val.txt','w')for i in list_index:
    name = total_xml[i][:-4]+'\n'if i in trainval:
        file_trainval.write(name)if i in train:
            file_train.write(name)else:
            file_val.write(name)else:
        file_test.write(name)

file_trainval.close()
file_train.close()
file_val.close()
file_test.close()

运行之后，会发现VOCData文件夹里面多了ImageSets文件夹

里面包括四个txt文件

txt文件存储了图片的序号

4、新建xml_to_yolo.py，用于将xml格式的标签转化为yolo格式的标签
对于简单的xml文件，用下列代码

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

sets =['train','val','test']
classes =["ship"]  # 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

def convert_annotation(image_id):
    in_file =open('D:/Project/yolov5Project/yolov5/VOCData/Annotations/%s.xml' % (image_id), encoding='UTF-8')
    out_file =open('D:/Project/yolov5Project/yolov5/VOCData/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/yolov5Project/yolov5/VOCData/labels/'):
        os.makedirs('D:/Project/yolov5Project/yolov5/VOCData/labels/')
    image_ids =open('D:/Project/yolov5Project/yolov5/VOCData/ImageSets/Main/%s.txt' %(image_set)).read().strip().split()if not os.path.exists('D:/Project/yolov5Project/yolov5/VOCData/dataSet_path/'):
        os.makedirs('D:/Project/yolov5Project/yolov5/VOCData/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/yolov5Project/yolov5/VOCData/images/%s.jpg\n' %(image_id))convert_annotation(image_id)
    list_file.close()

对于复杂的xml文件，用下列代码。
该程序的功能：将Class_ID属于aircraft_carrier列表的物体标记为类别0，将Class_ID属于warcraft列表的物体标记为类别1，将Class_ID属于aircraft_carrier列表的物体标记为类别2，将Class_ID属于merchant_ship列表的物体标记为类别3。然后转化为yolo格式标签。

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

Class_1_num =0
Class_2_num =0
Class_3_num =0
Class_4_num =0

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

aircraft_carrier =[100000005,100000006,100000012,100000013,100000031,100000032,100000033]
warcraft =[100000007,100000008,100000009,100000010,100000011,100000014,100000015,100000016,100000017,100000019,100000003,100000029]
merchant_ship =[100000018,100000022,100000024,100000018,100000025,100000026,100000030]
Submarine =[100000027]

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

def convert_annotation(image_id):
    global Class_1_num,Class_2_num,Class_3_num,Class_4_num
    in_file =open('./Annotations/%s.xml'%(image_id), encoding='UTF-8')
    out_file =open('./labels/%s.txt'%(image_id),'w')
    tree = ET.parse(in_file)
    root = tree.getroot()
    HRSC_Objects = root.find('HRSC_Objects')
    HRSC_Object = HRSC_Objects.find('HRSC_Object')if HRSC_Object!=None:for HRSC_Object in HRSC_Objects.iter('HRSC_Object'):
            Class_ID =int(HRSC_Object.find('Class_ID').text)if Class_ID in aircraft_carrier or Class_ID in warcraft or Class_ID in merchant_ship or Class_ID in Submarine:
                w =int(root.find('Img_SizeWidth').text)
                h =int(root.find('Img_SizeHeight').text)
                b =(float(HRSC_Object.find('box_xmin').text),float(HRSC_Object.find('box_xmax').text),float(HRSC_Object.find('box_ymin').text),float(HRSC_Object.find('box_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)if Class_ID in aircraft_carrier:
                    Class_1_num = Class_1_num+1
                    Class =0if Class_ID in warcraft:
                    Class_2_num = Class_2_num+1
                    Class =1if Class_ID in merchant_ship:
                    Class_3_num = Class_3_num+1
                    Class =2if Class_ID in Submarine:
                    Class_4_num = Class_4_num+1
                    Class =3
                out_file.write(str(Class)+" "+" ".join([str(a)for a in bb])+'\n')else:print('do not caculate'+str(Class_ID))continue

wd =getcwd()for image_set in sets:if not os.path.exists('./labels/'):
        os.makedirs('./labels/')
    image_ids =open('./ImageSets/Main/%s.txt'%(image_set)).read().strip().split()if not os.path.exists('./dataSet_path/'):
        os.makedirs('./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:print('目前的图片id为：'+image_id)
        list_file.write('D:/Project/yolov5Project/yolov5_three_classifications/VOCData/images/%s.bmp\n' %(image_id))convert_annotation(image_id)
    list_file.close()print(Class_1_num)print(Class_2_num)print(Class_3_num)print(Class_4_num)

运行后，会发现多了两个文件

dataSet_path文件中存储了训练集、测试集、验证集的绝对路径

labels存储了yolo格式的标签


5、进入data文件夹，新建mycov.yaml

myvoc.yaml的内容，其中，train与val为训练集与验证集列表的绝对路径

6、修改权重文件yolov5s.yaml

把类别改为我们设置的类别数量

7、开始训练
（1）训练

python train.py --weights weights/yolov5s.pt  --cfg models/yolov5s.yaml  --data data/myvoc.yaml --epoch 100--batch-size 16--img 640--device 0

（2）训练过程可视化

tensorboard --logdir=./runs

（3）测试训练出的网络模型

python detect.py --source data/images/000026.jpg --weights runs/train/exp10/weights/best.pt

一些常用命令：
（1）指定下载镜像源

pip install onnx -i http://pypi.douban.com/simple --trusted-host pypi.douban.com

（2）制作数据集教程

（3）mAP说明