1.在darknet目录下创建myData文件夹,目录结构如下,将labelimg工具标注好的图片和xml文件放到对应目录下:
myData
…JPEGImages#存放图像
…Annotations#存放图像对应的xml文件
…ImageSets/Main # 存放训练/验证图像的名字(格式如 000001.jpg或者000001),里面包括train.txt。这里给出的格式是: 000000,因为下面的代码中给出了图像的格式。
类似如下格式:
2.将自己的数据集图片拷贝到JPEGImages目录下。将数据集label文件拷贝到Annotations目录下。在myData下创建test.py,将下面代码拷贝进去运行,将生成四个文件:train.txt,val.txt,test.txt和trainval.txt。
import os
import random
trainval_percent = 0.1
train_percent = 0.9
xmlfilepath = 'Annotations'
txtsavepath = 'ImageSets\Main'
total_xml = os.listdir(xmlfilepath)
num = len(total_xml)
list = range(num)
tv = int(num * trainval_percent)
tr = int(tv * train_percent)
trainval = random.sample(list, tv)
train = random.sample(trainval, tr)
ftrainval = open('ImageSets/Main/trainval.txt', 'w')
ftest = open('ImageSets/Main/test.txt', 'w')
ftrain = open('ImageSets/Main/train.txt', 'w')
fval = open('ImageSets/Main/val.txt', 'w')
for i in list:
name = total_xml[i][:-4] + '\n'
if i in trainval:
ftrainval.write(name)
if i in train:
ftest.write(name)
else:
fval.write(name)
else:
ftrain.write(name)
ftrainval.close()
ftrain.close()
fval.close()
ftest.close()
3.将数据转换成darknet支持的格式
yolov3提供了将VOC数据集转为YOLO训练所需要的格式的代码,在scripts/voc_label.py文件中。这里提供一个修改版本的。在darknet文件夹下新建一个my_lables.py文件,内容如下(路径需要按照自己实际路径进行修改):
import xml.etree.ElementTree as ET
import pickle
import os
from os import listdir, getcwd
from os.path import join
#源代码sets=[('2012', 'train'), ('2012', 'val'), ('2007', 'train'), ('2007', 'val'), ('2007', 'test')]
# 训练模式下改为sets=[('myData', 'train'), ('myData', 'test')]
sets=[('myData', 'train')] # 改成自己建立的myData
classes = ["person", "foot", "face"] # 改成自己的类别
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(year, image_id):
in_file = open('myData/Annotations/%s.xml'%(image_id)) # 源代码VOCdevkit/VOC%s/Annotations/%s.xml
out_file = open('myData/labels/%s.txt'%(image_id), 'w') # 源代码VOCdevkit/VOC%s/labels/%s.txt
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('myData/labels/'): # 改成自己建立的myData
os.makedirs('myData/labels/')
image_ids = open('myData/ImageSets/Main/%s.txt'%(image_set)).read().strip().split()
list_file = open('myData/%s_%s.txt'%(year, image_set), 'w')
for image_id in image_ids:
list_file.write('%s/myData/JPEGImages/%s.jpg\n'%(wd, image_id))
convert_annotation(year, image_id)
list_file.close()
3.运行该脚本
python my_lables.py
4.修改darknet/cfg下的voc.data和yolov3-voc.cfg文件
为了保险起见,复制这两个文件,并分别重命名为my_data.data和yolov3-tiny-car.cfg
my_data.data内容:
classes= 1 ##改为自己的分类个数
##下面都改为自己的路径
train = darknet_pwd + /myData/myData_train.txt
names = darknet_pwd + /myData/myData.names #稍后需要创建这个文件
backup = /backup # 模型生成路径
yolov3-tiny-car.cfg的内容:
/yolo, 总共会搜出3个含有yolo的地方。
每个地方都必须要改2处, filters:3*(5+len(classes));
其中:classes: len(classes) = 1,这里以我的工程为例
filters = 24
classes = 1
可修改:random = 1:原来是1,显存小改为0。(是否要多尺度输出。)
修改成训练模式
因为是训练,所以注释Testing,打开Training,其中
batch=64 每batch个样本更新一次参数。
subdivisions=16 如果内存不够大,将batch分割为subdivisions个子batch,每个子batch的大小为batch/subdivisions。
6.设置anchor box
https://github.com/lars76/kmeans-anchor-boxes
下载kmeans.py
同时新增一个anchor_box.py文件
# -*- coding=utf-8 -*-
import glob
import os
import sys
import xml.etree.ElementTree as ET
import numpy as np
from kmeans import kmeans, avg_iou
# 根文件夹
ROOT_PATH = '/data/DataBase/YOLO_Data/V3_DATA/'
# 聚类的数目
CLUSTERS = 6
# 模型中图像的输入尺寸,默认是一样的
SIZE = 640
# 加载YOLO格式的标注数据
def load_dataset(path):
jpegimages = os.path.join(path, 'JPEGImages')
if not os.path.exists(jpegimages):
print('no JPEGImages folders, program abort')
sys.exit(0)
labels_txt = os.path.join(path, 'labels')
if not os.path.exists(labels_txt):
print('no labels folders, program abort')
sys.exit(0)
label_file = os.listdir(labels_txt)
print('label count: {}'.format(len(label_file)))
dataset = []
for label in label_file:
with open(os.path.join(labels_txt, label), 'r') as f:
txt_content = f.readlines()
for line in txt_content:
line_split = line.split(' ')
roi_with = float(line_split[len(line_split)-2])
roi_height = float(line_split[len(line_split)-1])
if roi_with == 0 or roi_height == 0:
continue
dataset.append([roi_with, roi_height])
# print([roi_with, roi_height])
return np.array(dataset)
data = load_dataset(ROOT_PATH)
out = kmeans(data, k=CLUSTERS)
print(out)
print("Accuracy: {:.2f}%".format(avg_iou(data, out) * 100))
print("Boxes:\n {}-{}".format(out[:, 0] * SIZE, out[:, 1] * SIZE))
ratios = np.around(out[:, 0] / out[:, 1], decimals=2).tolist()
print("Ratios:\n {}".format(sorted(ratios)))
经过运行之后得到一组如下数据:
[[0.21203704 0.02708333]
[0.34351852 0.09375 ]
[0.35185185 0.06388889]
[0.29513889 0.06597222]
[0.24652778 0.06597222]
[0.24861111 0.05347222]]
Accuracy: 89.58%
Boxes:
[135.7037037 219.85185185 225.18518519 188.88888889 157.77777778
159.11111111]-[17.33333333 60. 40.88888889 42.22222222 42.22222222 34.22222222]
其中的Boxes就是得到的anchor参数,以上面给出的计算结果为例,最后的anchor参数设置为
anchors = 135,17, 219,60, 225,40, 188,42, 157,42, 159,34
6.训练
在myData文件夹下新建myData.names文件(标签文件)
car
foot
people
生成预训练权重(tiny)
wget https://pjreddie.com/media/files/yolov3-tiny.weights(如果有可以不下载)
./darknet partial ./cfg/yolov3-tiny.cfg ./yolov3-tiny.weights ./yolov3-tiny.conv.15 15
或者直接下载预训练权重
wget https://pjreddie.com/media/files/darknet53.conv.74
开始训练(我用的是tiny模型)
./darknet detector train ./cfg/mydata.data ./cfg/yolov3-tiny-car.cfg ./mydata/yolov3-tiny.conv.15
或者指定gpu训练,默认使用gpu0
./darknet detector train cfg/my_data.data cfg/yolov3-tiny-car.cfg darknet53.conv.74 -gups 0,1,2,3
从停止处重新训练
./darknet detector train cfg/my_data.data cfg/yolov3-tiny-car.cfg darknet53.conv.74 -gups 0,1,2,3 backup/my_yolov3.backup -gpus 0,1,2,3
测试
./darknet detector test cfg/mydata.data cfg/yolov3-tiny-car.cfg backup/yolov3-tiny-mydata_900.weights 2.jpg
完成