一:yolov5的安装
yolov5的下载、安装及其环境配置可以参考吴青峰老师的文章。【精选】yolov5/yolov3训练自己的数据集(超详细,细的有点烦)_yolo数据集_吴青峰-CDP的博客-CSDN博客
二:准备数据集并标注
1.可通过以下代码收集需求的图片
#导入相应的库
import requests
import re
import os
#获取网站的源代码
def get_html(url,headers,params):
response=requests.get(url,headers=headers,params=params)
#设置源代码的编码方式
response.encoding="utf-8"
if response.status_code == 200:
return response.text
else:
print("网站源码获取错误")
#提取图片的源地址
def parse_pic_url(html):
result = re.findall('thumbURL":"(.*?)"',html,re.S)
return result
#获取图片的二进制源码
def get_pic_content(url):
response = requests.get(url)
return response.content
#保存图片
def save_pic(fold_name,content,pic_name):
with open(fold_name+"/"+str(pic_name)+".jpg","wb") as f:
f.write(content)
f.close()
#定义一个新建文件夹程序
def creat_fold(fold_name):
#加异常处理
try:
os.mkdir(fold_name)
except:
print("文件夹已存在")
#定义main函数调用get_html函数
def main():
try:
#输入文件夹的名字
fold_name = input("请输入你要抓取的图片名字: ")
#输入要抓取的图片页数
page_num = input("请输入要抓取多少页?(0,1,2,3...)")
#调用函数,创建文件夹
creat_fold(fold_name)
# 定义图片名字
pic_name = 0
#构建循环,控制页面
for i in range(int(page_num)):
try:
url="https://image.baidu.com/search/acjson?tn=resultjson_com&logid=11578134858473490309&ipn=rj&ct=201326592&is=&fp=result&fr=&word=%E5%A4%A7%E7%86%8A%E7%8C%AB&queryWord=%E5%A4%A7%E7%86%8A%E7%8C%AB&cl=2&lm=-1&ie=utf-8&oe=utf-8&adpicid=&st=&z=&ic=&hd=&latest=©right=&s=&se=&tab=&width=&height=&face=&istype=&qc=&nc=1&expermode=&nojc=&isAsync=&pn=30&rn=30&gsm=1e&1695863823504="
headers={
"Accept":"text/plain, */*; q=0.01",
"Accept-Encoding":"gzip, deflate, br",
"Accept-Language":"zh-CN,zh;q=0.9",
"Connection":"keep-alive",
"Cookie":"BIDUPSID=C5F32270171263FD60A06574D820568E; PSTM=1663574847; BAIDUID=C5F32270171263FDD54F54A9E950CE44:SL=0:NR=10:FG=1; indexPageSugList=%5B%22%E5%92%96%E5%95%A1%E5%85%AC%E5%8F%B8%E5%A4%96%E8%A1%A8%E5%9B%BE%E7%89%87%22%2C%22%E5%92%96%E5%95%A1%E5%85%AC%E5%8F%B8%E5%9B%BE%E7%89%87%22%2C%22%E5%92%96%E5%95%A1%E5%AE%A3%E4%BC%A0%E7%85%A7%E6%A8%AA%E5%B1%8F%22%2C%22%E6%A3%95%E8%89%B2%E8%B0%83%E5%92%96%E5%95%A1%E5%AE%A3%E4%BC%A0%E7%85%A7%E7%89%87%E6%A8%AA%E5%B1%8F%22%2C%22%E5%92%96%E5%95%A1%E5%AE%A3%E4%BC%A0%E7%85%A7%E7%89%87%E6%A3%95%E8%89%B2%E8%B0%83%E6%A8%AA%E5%B1%8F%22%2C%22%E5%89%8D%E7%AB%AF%E7%85%A7%E7%89%87%E5%A6%82%E4%BD%95%E5%BE%AA%E7%8E%AF%E6%92%AD%E6%94%BE%22%2C%22%E5%92%96%E5%95%A1%E5%AE%A3%E4%BC%A0%E7%85%A7%E7%89%87%22%2C%22%E5%92%96%E5%95%A1%E7%85%A7%E7%89%87%E5%8F%8A%E5%90%8D%E5%AD%97%22%2C%22%E8%B4%AD%E7%89%A9%E9%A1%B5%E9%9D%A2%E6%B6%88%E8%B4%B9%E8%AE%B0%E5%BD%95%E8%AE%BE%E8%AE%A1%22%5D; BAIDUID_BFESS=C5F32270171263FDD54F54A9E950CE44:SL=0:NR=10:FG=1; ZFY=6f4:B7Z:AO5IlHn7wNjMT7JLCppMwmRurc9Me:A:Art5uEU:C; BA_HECTOR=0la085a4008ga1052g20ag0j1ih799e1p; BDORZ=B490B5EBF6F3CD402E515D22BCDA1598; PSINO=1; H_PS_PSSID=39309_39354_39399_39396_39352_39407_39097_39411_39436_39358_39450_39233_39403_26350_39425; delPer=0; BDRCVFR[dG2JNJb_ajR]=mk3SLVN4HKm; userFrom=www.baidu.com; BDRCVFR[-pGxjrCMryR]=mk3SLVN4HKm",
"Host":"image.baidu.com",
"Referer":"https://image.baidu.com/search/index?tn=baiduimage&ps=1&ct=201326592&lm=-1&cl=2&nc=1&ie=utf-8&dyTabStr=MTEsMCwxLDYsMyw0LDUsMiw4LDcsOQ%3D%3D&word=%E5%A4%A7%E7%86%8A%E7%8C%AB",
"Sec-Ch-Ua":'"Chromium";v="116", "Not)A;Brand";v="24", "Google Chrome";v="116"',
"Sec-Ch-Ua-Mobile":"?0",
"Sec-Ch-Ua-Platform":'"Windows"',
"Sec-Fetch-Dest":"empty",
"Sec-Fetch-Mode":"cors",
"Sec-Fetch-Site":"same-origin",
"User-Agent":"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/116.0.0.0 Safari/537.36",
"X-Requested-With":"XMLHttpRequest"
}
params={
"tn": "resultjson_com",
"logid": "11325976662758874031",
"ipn":"rj",
"ct": "201326592",
"fp": "result",
"word": fold_name,
"queryWord": fold_name,
"cl": "2",
"lm": "-1",
"ie": "utf-8",
"oe": "utf-8",
"nc": "1",
"pn": str(int(i+1)*30),
"rn": "30",
"gsm": "1e"
}
html = get_html(url,headers,params)
#print()
result = parse_pic_url(html)
#使用for循环遍历列表
for item in result:
#调用函数,获取图片的二进制源码
pic_content = get_pic_content(item)
#调用函数保存图片
save_pic(fold_name,pic_content,pic_name)
pic_name += 1
print("正在保存第 "+str(pic_name)+" 张图片")
except:
print("抓取第 "+str(i)+" 页错误")
except:
print("数据抓取异常......")
#执行main函数
if __name__ == '__main__':
main()
运行效果如下,可根据自己实际情况填写:
2.将图片放入yolov5-master中的image文件夹中,在data文件夹里新建一个dataset文件夹、一个labels文件夹用于后续使用。dataset用来存放打好标签的数据集,labels用于存放未打标签的。
3.进行数据的标注,首先打开终端输入 pip install labelimg -i https://pypi.tuna.tsinghua.edu .cn/sim ple安装标注工具labelimg。
4.打开labelimg开始进行标注
注意:
标注过程中也可使用快捷键。W可进行标注,S表示上一张,D表示下一张
5.对所有的图片打好标签后,dataset里会生成多个xml文件
三:数据预处理与源代码修改
接下来我们来对数据进行预处理操作:
首先需要在yolov5工程目录下新建两个.py程序,split.py 和 xml_to_txt.py
split.py代码如下:
import os
import random
import argparse
parser = argparse.ArgumentParser()
parser.add_argument('--xml_path', default='data/dataset', type=str, help='input xml label path')
parser.add_argument('--txt_path', default='data/labels', type=str, help='output txt label path')
opt = parser.parse_args()
trainval_percent = 1.0
train_percent = 0.8
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()
然后运行此代码,运行完成后的结果如下图所示,labels标签下会出现4个txt文件:
xml_to_txt.py 代码如下:
import xml.etree.ElementTree as ET
from tqdm import tqdm
import os
from os import getcwd
sets = ['train', 'val', 'test']
classes = ['hand'] # 这里改为你要训练的标签,否则会报错。比如你要识别“hand”,那这里就改为hand
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):
# try:
in_file = open('data/dataset/%s.xml' % (image_id), encoding='utf-8')
out_file = open('data/labels/%s.txt' % (image_id), 'w', encoding='utf-8')
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 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')
# except Exception as e:
# print(e, image_id)
wd = getcwd()
for image_set in sets:
if not os.path.exists('data/labels/'):
os.makedirs('data/labels/')
image_ids = open('data/labels/%s.txt' %
(image_set)).read().strip().split()
list_file = open('data/%s.txt' % (image_set), 'w')
for image_id in tqdm(image_ids):
list_file.write('data/images/%s.jpg\n' % (image_id))
convert_annotation(image_id)
list_file.close()
以上代码中classes = [“你要训练的标签名称”],在本代码中我要训练的数据是识别hand,所以我这里改成了“hand”。大家根据自己的情况进行修改。运行此代码,运行后的结果如下图所示:
运行之后,data下的labels文件夹里看会有txt文件。
接下来在data文件夹中新建myvoc.yaml文件
代码如此就改写完了,接下来可以开始训练了
四:训练自己的数据集
打开终端,打开yolov5,输入下面这段代码:
python train.py --epoch 300 --batch 3 --data ./data/myvoc.yaml --cfg ./models/yolov5s.yaml --weight ./weights/yolov5s.pt --workers 0
出现此画面代表数据开始训练。
最后训练结束后会出现一个last.py文件和best.py文件
在文件夹里找到它的位置,直接复制到weights中,
这样模型训练就完成了。
五:检测
在训练完的结果后输入命令:
python detect.py --weight ./weights/last.pt --source 0
摄像头出现后便可以开始识别数据了。