准备工作:
参考链接:https://pjreddie.com/darknet/yolo/
文件下载:yolov4.weights
、yolov4.conv.137
一、源码测试
源码链接:https://github.com/AlexeyAB/darknet
1.1 在Ubuntu编译:
a:在cmakelist.txt内修改配置文件,修改为已有文件版本
GPU=1使用CUDA进行构建以通过使用GPU加速(CUDA应该在中/usr/local/cuda)
CUDNN=1使用cuDNN v5-v7进行构建,以通过使用GPU加速培训(cuDNN应该在中/usr/local/cudnn)
CUDNN_HALF=1 为Tensor Core构建(在Titan V / Tesla V100 / DGX-2及更高版本上)加速检测3倍,训练2倍
OPENCV=1 使用OpenCV 4.x / 3.x / 2.4.x进行构建-允许检测来自网络摄像机或网络摄像机的视频文件和视频流
DEBUG=1 调试Yolo版本
OPENMP=1 使用OpenMP支持进行构建以通过使用多核CPU来加速Yolo
LIBSO=1生成一个库darknet.so和uselib使用该库的二进制可运行文件
b:在终端运行./build.sh
c:编译,使用make
测试代码:
在Linux
上./darknet
,在根目录中找到可执行文件,而在Windows
上,在目录中找到可执行文件。\build\darknet\x64
# 1 Yolo v4 COCO-图片:
./darknet detector test cfg/coco.data cfg/yolov4.cfg yolov4.weights -thresh 0.25
# 2 输出对象的坐标:
./darknet detector test cfg/coco.data yolov4.cfg yolov4.weights -ext_output dog.jpg
# 3 Yolo v4 COCO-视频:
./darknet detector demo cfg/coco.data cfg/yolov4.cfg yolov4.weights -ext_output test.mp4
# 4 Yolo v4-保存结果视频文件res.avi:
./darknet detector demo cfg/coco.data cfg/yolov4.cfg yolov4.weights test.mp4 -out_filename res.avi
# 5 要计算锚点:
./darknet detector calc_anchors data/obj.data -num_of_clusters 9 -width 416 -height 416
# 5 要检查准确性mAP @ IoU = 50:
./darknet detector map data/obj.data yolo-obj.cfg backup\yolo-obj_7000.weights
二、准备数据集
darknet-master\build\darknet\x64文件夹下新建了一个myData文件夹,存放要训练的数据集。
1.1 在文件夹内放置图片,生成训练文件
myData
......JPEGImages #存放图像
......Annotations #存放图像对应的xml文件
......ImageSets/Main #存放训练/存放train.txt/val.txt/test.txt/trainval.txt文件
......test.py #生成train.txt/val.txt/test.txt/trainval.txt文件
darknet-master\build\darknet\x64\myData文件夹下运行test.py
test.py
代码如下
import os
import random
xmlfilepath=r"Annotations"
saveBasePath=r"\\ImageSets\\Main"
trainval_percent=1
train_percent=0.1
'''temp_xml = os.listdir(xmlfilepath)
total_xml = []
for xml in temp_xml:
if xml.endswith(".xml"):
total_xml.append(xml)
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)
print("train and val size",tv)
print("traub suze",tr)
ftrainval = open(os.path.join(saveBasePath,'trainval.txt'), 'w')
ftest = open(os.path.join(saveBasePath,'test.txt'), 'w')
ftrain = open(os.path.join(saveBasePath,'train.txt'), 'w')
fval = open(os.path.join(saveBasePath,'val.txt'), 'w')
'''
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:
ftrain.write(name)
else:
fval.write(name)
else:
ftest.write(name)
ftrainval.close()
ftrain.close()
fval.close()
ftest .close()
1.2 在darknet-master\build\darknet\x64文件夹下新建my_labels.py
运行该脚本my_lables.py
会在./myData
目录下生成一个labels
文件夹一个txt文件(myData_train.txt
)(内容是: 类别的编码和目标的相对位置)。
my_labels.py
代码如下:
import xml.etree.ElementTree as ET
import pickle
import os
from os import listdir, getcwd
from os.path import join
sets=[('myData', 'train'), ('myData', 'val'), ('myData', 'train'), ('myData', 'val'), ('myData', 'test')]
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))
out_file = open('myData/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
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()
1.3 在my_data
文件夹新建文件myData.names
文件
# 填入数据集名称
person
cat
1.4 在myData
文件夹下新建weights
文件,用于保存训练产生的权重文件
1.5修改配置文件
在darknet/cfg 下新建my_data.data和my_yolov4.cfg文件
a:修改my_data.data:
classes= 2 #改为自己的分类个数
##下面都改为自己的路径
train = ../darknet/myData/myData_train.txt
valid =../darknet/myData/myData_test.txt
names = ../darknet/myData/myData.names
backup = ../darknet/myData/weights
b:将yolov4.cfg
文件内容复制到my_yolov4.cfg
文件并修改my_yolov4.cfg
文件
1):
max_batches
改成合适的数值,一般为classes*2000
steps
分别取max_batches
的80%和90%
2)
修改3个含有yolo部位的地方,修改filters
、classes
filters
:3*( 5 + len (classes) )
1.6 训练前操作:
a:将my_yolov4.cfg
文件中改成Training
模式
b:根据电脑配置要求修改batch
和subdivisions
的参数
c:根据显存修改my_yolov4.cfg
中random
原来是1,显存小改为0
三、数据训练
./darknet detector train cfg/my_data.data cfg/my_yolov4.cfg yolov4.conv.137
# 训练时显示map
./darknet detector train cfg/my_data.data cfg/my_yolov4.cfg yolov4.conv.137 -map
# 在无显示工具或不希望在图上显示,可以直接保存
./darknet detector train cfg/my_data.data cfg/my_yolov4.cfg yolov4.conv.137 -map -dont_show
# 指定gpu训练,默认使用gpu0(查看GPU情况,`nvidia-smi`)
./darknet detector train cfg/my_data.data cfg/my_yolov4.cfg yolov4.conv.137 -gups 0,1,2,3
# 训练过程中保存训练日志xxx.log
./darknet detector train cfg/my_data.data cfg/my_yolov4.cfg yolov4.conv.137 | tee train_yolov4.log
# 断点继续训练
./darknet detector train cfg/my_data.data cfg/my_yolov4.cfg myData/weights/my_yolov4.backup | tee new_train_yolov4.log
4、数据集检测
同上源码测试,修改权重文件。