yolov3训练自己的数据

最新推荐文章于 2024-06-05 09:51:16 发布

知识在于分享

最新推荐文章于 2024-06-05 09:51:16 发布

阅读量558

点赞数

分类专栏：深度学习

本文链接：https://blog.csdn.net/baidu_40840693/article/details/95065151

版权

深度学习专栏收录该内容

255 篇文章 18 订阅

订阅专栏

这批数据，名字带空格，需要先处理一下：

#-*-coding:utf-8-*-

import os

class BatchRename():
    '''
    批量重命名文件夹中的图片文件

    '''
    def __init__(self):
        self.path = r'./JPEGImages'  #表示需要命名处理的文件夹

    def rename(self):
        filelist = os.listdir(self.path) #获取文件路径
        total_num = len(filelist) #获取文件长度（个数）
        i = 1  #表示文件的命名是从1开始的
        for item in filelist:
            if item.endswith('.jpg'):  #初始的图片的格式为jpg格式的（或者源文件是png格式及其他格式，后面的转换格式就可以调整为自己需要的格式即可）
                src = os.path.join(os.path.abspath(self.path), item)
                # dst = os.path.join(os.path.abspath(self.path), '2_' + str(i) + '.jpg')#处理后的格式也为jpg格式的，当然这里可以改成png格式
                # dst = os.path.join(os.path.abspath(self.path), '002_' + str(i) + '.jpg')
                dst = os.path.join(os.path.abspath(self.path), '0000' + format(str(i), '0>3s') + '.jpg')    #这种情况下的命名格式为0000000.jpg形式，可以自主定义想要的格式
                try:
                    os.rename(src, dst)
                    print('converting %s to %s ...' % (src, dst))
                    i = i + 1
                    src = src.replace("JPEGImages","Annotations")
                    src = src.replace(".jpg",".xml")
                    dst = dst.replace("JPEGImages","Annotations")
                    dst = dst.replace(".jpg",".xml")
                    os.rename(src, dst)
                    print('converting %s to %s ...' % (src, dst))
                except:
                    continue
        print('total %d to rename & converted %d jpgs' % (total_num, i))

if __name__ == '__main__':
    demo = BatchRename()
    demo.rename()

ImageSets/Main

生成数据的train.txt

#-*-coding:utf-8-*-
import os
import argparse

# description参数可以用于插入描述脚本用途的信息，可以为空
# 添加--verbose标签，标签别名可以为-v，这里action的意思是当读取的参数中出现--verbose/-v的时候
# 参数字典的verbose建对应的值为True，而help参数用于描述--verbose参数的用途或意义。
# 将变量以标签-值的字典形式存入args字典

def main(args):
    '''main '''
    flist = open('train.txt', 'w')
    all_image = os.listdir('./' + args.file)
    num = len(all_image)
    number = 0
    for im in all_image:
        if(im.split('.')[-1] != 'jpg'):
            continue
        number += 1
	im = im.split('.')[0]
        print number, "  /  ", num
        flist.write(im + '\n')
    flist.close()

# https://www.cnblogs.com/arkenstone/p/6250782.html
def parse_args():
    '''parse args'''
    parser = argparse.ArgumentParser()
    parser.add_argument('--file', '-f', default='source')
    return parser.parse_args()

if __name__ == '__main__':
    print False
    print True
    print("False %s" % (False))
    print("True %s" % (True))
    main(parse_args())

发现这批数据的XML格式与VOC格式不同：类别为1个

<?xml version="1.0" ?>
<doc>
	<path>E:\4\bbk2\1 (219).jpg</path>
	<outputs>
		<object>
			<item>
				<name>book</name>
				<bndbox>
					<xmin>576</xmin>
					<ymin>748</ymin>
					<xmax>1988</xmax>
					<ymax>2244</ymax>
				</bndbox>
			</item>
		</object>
	</outputs>
	<time_labeled>1562292940774</time_labeled>
	<labeled>true</labeled>
	<size>
		<width>2448</width>
		<height>3264</height>
		<depth>3</depth>
	</size>
</doc>

<?xml version="1.0" ?>
<doc>
	<path>E:\4\bbk2\1 (214).jpg</path>
	<outputs></outputs>
	<time_labeled>0</time_labeled>
	<labeled>false</labeled>
</doc>

处理方式改变一下：

import xml.etree.ElementTree as ET
import pickle
import os
from os import listdir, getcwd
from os.path import join

sets=[('2019', 'train')]


def convert(size, box):
    dw = 1./size[0]
    dh = 1./size[1]
    x = (box[0] + box[1])/2.0
    y = (box[2] + box[3])/2.0
    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(year, image_id):
    in_file = open('VOCdevkit/VOC%s/Annotations/%s.xml'%(year, image_id))
    tree=ET.parse(in_file)
    root = tree.getroot()
    size = root.find('size')
    if size==None:
        return None
    w = int(size.find('width').text)
    h = int(size.find('height').text)

    outputs = root.find('outputs')
    objects = outputs.find('object')

    out_file = open('VOCdevkit/VOC%s/labels/%s.txt'%(year, image_id), 'w')
    for obj in objects:
        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))
        bb = convert((w,h), b)
        out_file.write(str(0) + " " + " ".join([str(a) for a in bb]) + '\n')
    return 1

wd = getcwd()

for year, image_set in sets:
    if not os.path.exists('VOCdevkit/VOC%s/labels/'%(year)):
        os.makedirs('VOCdevkit/VOC%s/labels/'%(year))
    image_ids = open('VOCdevkit/VOC%s/ImageSets/Main/%s.txt'%(year, image_set))
    list_file = open('%s_%s.txt'%(year, image_set), 'w')
    for image_id in image_ids:
        print image_id
        image_id = image_id.replace('\n','')
        flag = convert_annotation(year, image_id)
        if flag == 1:
            list_file.write('%s/VOCdevkit/VOC%s/JPEGImages/%s.jpg\n'%(wd, year, image_id))
    list_file.close()

顺便贴一下原始的python：

import xml.etree.ElementTree as ET
import pickle
import os
from os import listdir, getcwd
from os.path import join

sets=[('2012', 'train'), ('2012', 'val'), ('2007', 'train'), ('2007', 'val'), ('2007', 'test')]

classes = ["aeroplane", "bicycle", "bird", "boat", "bottle", "bus", "car", "cat", "chair", "cow", "diningtable", "dog", "horse", "motorbike", "person", "pottedplant", "sheep", "sofa", "train", "tvmonitor"]


def convert(size, box):
    dw = 1./size[0]
    dh = 1./size[1]
    x = (box[0] + box[1])/2.0
    y = (box[2] + box[3])/2.0
    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(year, image_id):
    in_file = open('VOCdevkit/VOC%s/Annotations/%s.xml'%(year, image_id))
    out_file = open('VOCdevkit/VOC%s/labels/%s.txt'%(year, 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
        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))
        bb = convert((w,h), b)
        out_file.write(str(cls_id) + " " + " ".join([str(a) for a in bb]) + '\n')

wd = getcwd()

for year, image_set in sets:
    if not os.path.exists('VOCdevkit/VOC%s/labels/'%(year)):
        os.makedirs('VOCdevkit/VOC%s/labels/'%(year))
    image_ids = open('VOCdevkit/VOC%s/ImageSets/Main/%s.txt'%(year, image_set)).read().strip().split()
    list_file = open('%s_%s.txt'%(year, image_set), 'w')
    for image_id in image_ids:
        list_file.write('%s/VOCdevkit/VOC%s/JPEGImages/%s.jpg\n'%(wd, year, image_id))
        convert_annotation(year, image_id)
    list_file.close()

重新用label的文件夹生成train.txt

因为JPEGImages中有部分不符合标准的xml：改一个地方就行：

if(im.split('.')[-1] != 'txt'):

#-*-coding:utf-8-*-
import os
import argparse

# description参数可以用于插入描述脚本用途的信息，可以为空
# 添加--verbose标签，标签别名可以为-v，这里action的意思是当读取的参数中出现--verbose/-v的时候
# 参数字典的verbose建对应的值为True，而help参数用于描述--verbose参数的用途或意义。
# 将变量以标签-值的字典形式存入args字典

def main(args):
    '''main '''
    flist = open('train.txt', 'w')
    all_image = os.listdir('./' + args.file)
    num = len(all_image)
    number = 0
    for im in all_image:
        if(im.split('.')[-1] != 'txt'):
            continue
        number += 1
	im = im.split('.')[0]
        print number, "  /  ", num
        flist.write(im + '\n')
    flist.close()

# https://www.cnblogs.com/arkenstone/p/6250782.html
def parse_args():
    '''parse args'''
    parser = argparse.ArgumentParser()
    parser.add_argument('--file', '-f', default='source')
    return parser.parse_args()

if __name__ == '__main__':
    print False
    print True
    print("False %s" % (False))
    print("True %s" % (True))
    main(parse_args())

接着：

在

darknet$ tree
.
├── data
│   ├── yolov3-tiny.name
│   └── yolov3.names
├── bad.list
├── cfg
│   ├── yolov3.cfg
│   ├── yolov3.data
│   ├── yolov3-tiny.cfg
│   └── yolov3-tiny.data

我们拿tiny举例子：

yolov3-tiny.name

book

yolov3-tiny.data

classes= 1
train  = /home/boyun/data/VOCdevkit/VOC2019/train.txt
#valid  = coco_testdev
#valid = data/coco_val_5k.list
names = data/yolov3.names
backup = ./backup/
#eval=coco

原始：

yolov3-tiny.cfg

[convolutional]
size=1
stride=1
pad=1
filters=255
activation=linear



[yolo]
mask = 3,4,5
anchors = 10,14,  23,27,  37,58,  81,82,  135,169,  344,319
classes=80
num=6
jitter=.3
ignore_thresh = .7
truth_thresh = 1
random=1

更改：

[convolutional]
size=1
stride=1
pad=1
filters=18
activation=linear



[yolo]
mask = 3,4,5
anchors = 10,14,  23,27,  37,58,  81,82,  135,169,  344,319
classes=1
num=6
jitter=.3
ignore_thresh = .7
truth_thresh = 1
random=1

./darknet detector train cfg/yolov3.data cfg/yolov3.cfg darknet53.conv.74 2>1 | tee ./backup/log.txt

./darknet detector train cfg/yolov3-tiny.data cfg/yolov3-tiny.cfg yolov3-tiny.weights 2>1 | tee ./backup/log.txt

我们这里不能使用tiny的预训练权重，因为改了一些参数

但是：

darknet53.conv.74

还是可以用的

./darknet detector test cfg/yolov3.data cfg/yolov3.cfg backup/yolov3_3000.weights ./data/one_2019_05_29_00000002.jpg

./darknet detector test cfg/yolov3-tiny.data cfg/yolov3-tiny.cfg backup/yolov3-tiny_3000.weights ./data/one_2019_05_29_00000002.jpg

知识在于分享

关注

0
点赞
踩
1

收藏

觉得还不错? 一键收藏
1
评论
yolov3训练自己的数据

这批数据，名字带空格，需要先处理一下：#-*-coding:utf-8-*-import osclass BatchRename(): ''' 批量重命名文件夹中的图片文件 ''' def __init__(self): self.path = r'./JPEGImages' #表示需要命名处理的文件夹 def renam...
复制链接

扫一扫