文章参考链接地址:https://pjreddie.com/darknet/train-cifar/
由于实际的业务数据的要求,需要得到与之匹配的预训练模型,本文将使用YOLO的网络进行预训练,得到适合自己的分类器。这篇文章主要介绍如何利用Darknet框架训练分类器。
安装Darknet
若没有安装Darknet,需首先执行:
git clone https://github.com/pjreddie/darknet
cd darknet
make
- 1
- 2
- 3
- 4
获取数据集
可直接下载图片格式的数据集
将数据集放至darknet主目录下的data文件夹下,执行:
cd data
wget https://pjreddie.com/media/files/cifar.tgz
tar xzf cifar.tgz
- 1
- 2
- 3
- 4
查看cifar下的文件内容:
ls cifar
- 1
- 2
cifar包含两个文件夹:训练数据集(train)和测试集(test),以及一个标签文件labels.txt,可通过命令:
cat cifar/labels.txt
- 1
- 2
查看要学习的物体类别。
准备训练(测试)数据
首先需要生成训练和测试数据的路径文件train.txt,test.txt,执行下述命令:
cd cifar
find `pwd`/train -name \*.png > train.list
find `pwd`/test -name \*.png > test.list
cd ../..
- 1
- 2
- 3
- 4
- 5
制作数据集参数文件 ,在cfg/ 路径下新建文件 cfg/cifar.data,文件内容如下:
classes=10 //10个不同类别
train = data/cifar/train.list //训练数据路径
valid = data/cifar/test.list //验证数据路径
labels = data/cifar/labels.txt //标签文件
backup = backup/ //保存中间训练结果的路径
top=2 //测试时计算top_k的准确率
- 1
- 2
- 3
- 4
- 5
- 6
- 7
制作网络模型文件:
[net]
batch=64
subdivisions=8
width=224
height=224
channels=3
momentum=0.9
decay=0.0005
angle=0
saturation = .75
exposure = .75
hue=.1
#learning_rate=0.001
#max_batches = 120000
#policy=steps
#steps=-1,100,80000,100000
#scales=.1,10,.1,.1
learning_rate=0.001
max_batches = 80000
policy=poly
power=4
momentum=0.9
decay=0.0005
[convolutional]
batch_normalize=1
filters=8
size=3
stride=1
pad=1
activation=leaky
[maxpool]
size=2
stride=2
[convolutional]
batch_normalize=1
filters=16
size=3
stride=1
pad=1
activation=leaky
[maxpool]
size=2
stride=2
[convolutional]
batch_normalize=1
filters=32
size=3
stride=1
pad=1
activation=leaky
[maxpool]
size=2
stride=2
[convolutional]
batch_normalize=1
filters=64
size=3
stride=1
pad=1
activation=leaky
[maxpool]
size=2
stride=2
[convolutional]
batch_normalize=1
filters=128
size=3
stride=1
pad=1
activation=leaky
[maxpool]
size=2
stride=2
[convolutional]
batch_normalize=1
filters=256
size=3
stride=1
pad=1
activation=leaky
[maxpool]
size=2
stride=1
[convolutional]
batch_normalize=1
filters=512
size=3
stride=1
pad=1
activation=leaky
###########
[convolutional]
batch_normalize=1
size=3
stride=1
pad=1
filters=512
activation=leaky
[convolutional]
size=1
stride=1
pad=1
filters=10
activation=leaky
[avgpool]
[softmax]
groups=1
[cost]
type=sse
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10
- 11
- 12
- 13
- 14
- 15
- 16
- 17
- 18
- 19
- 20
- 21
- 22
- 23
- 24
- 25
- 26
- 27
- 28
- 29
- 30
- 31
- 32
- 33
- 34
- 35
- 36
- 37
- 38
- 39
- 40
- 41
- 42
- 43
- 44
- 45
- 46
- 47
- 48
- 49
- 50
- 51
- 52
- 53
- 54
- 55
- 56
- 57
- 58
- 59
- 60
- 61
- 62
- 63
- 64
- 65
- 66
- 67
- 68
- 69
- 70
- 71
- 72
- 73
- 74
- 75
- 76
- 77
- 78
- 79
- 80
- 81
- 82
- 83
- 84
- 85
- 86
- 87
- 88
- 89
- 90
- 91
- 92
- 93
- 94
- 95
- 96
- 97
- 98
- 99
- 100
- 101
- 102
- 103
- 104
- 105
- 106
- 107
- 108
- 109
- 110
- 111
- 112
- 113
- 114
- 115
- 116
- 117
- 118
- 119
- 120
- 121
- 122
- 123
- 124
- 125
- 126
- 127
- 128
- 129
- 130
- 131
训练模型
进入darknet根目录:
./darknet classifier train cfg/cifar.data cfg/my_cifar.cfg
- 1
- 2
若训练中断,基于已有的训练模型接着训练执行如下命令:
./darknet classifier train cfg/cifar.data cfg/my_cifar.cfg backup/my_cifar.backup
- 1
- 2
训练过程如下: