darknet classifier training alexnet

Create a profile

cfg under the darknet directory file directory contains the network .cfg, .namethe file indicates that the file needs to find a training model. file name as follows

# 分类类别数
classes=5
# 训练图像数据的完整路径列表，如/media/liushuai/新加卷/pocamen/dataset/bulbasaur/00000000.jpg
train=/media/liushuai/新加卷/pocamen/darknet_file/train.list 
# 验证文件的完整路径列表
valid=/media/liushuai/新加卷/pocamen/darknet_file/val.list
# 标签文件的名称如：bulbasaur（多目标则分行显示）
labels=/media/liushuai/新加卷/pocamen/darknet_file/labels.txt
# 权重文件的存储路径
backup=/media/liushuai/新加卷/pocamen/darknet_file/backup
# top2损失
top=2

cfg file is a file structure of the network, typically to modify the output of the original network structure and the loss function is added:

[net]
# Training
batch=128
subdivisions=1
# Testing
# batch=1
# subdivisions=1
height=227
width=227
channels=3
momentum=0.9
decay=0.0005
max_crop=256

learning_rate=0.01
policy=poly
power=4
max_batches=800000

angle=7
hue = .1
saturation=.75
exposure=.75
aspect=.75

[convolutional]
filters=96
size=11
stride=4
pad=0
activation=relu

[maxpool]
size=3
stride=2
padding=0

[convolutional] #自带的alaxnet这里的pad=1，按照pad=1计算输出结果应该输出错误，但是pad=2输出和pad=1输出一样，
filters=256
size=5
stride=1
pad=2
activation=relu

[maxpool]
size=3
stride=2
padding=0

[convolutional]
filters=384
size=3
stride=1
pad=1
activation=relu

[convolutional]
filters=384
size=3
stride=1
pad=1
activation=relu

[convolutional]
filters=256
size=3
stride=1
pad=1
activation=relu

[maxpool]
size=3
stride=2
padding=0

[connected]
output=4096
activation=relu

[dropout]
probability=.5

[connected]
output=4096
activation=relu

[dropout]
probability=.5

[connected]
output=5
activation=linear

[softmax]
groups=1

[cost]
type=sse #原始的alexnet网络结构没有定义这里

AlexNet network structure is as follows:

The training process is as follows:

./darknet classifier train cfg/pocman.data cfg/alexnet_pocaman.cfg

Forecasting process:

./darknet classifier predict cfg/pocman.data cfg/alexnet_pocaman.cfg /media/liushuai/新加卷/pocamen/darknet_file/backup/alexnet_pocaman_64.weights /media/liushuai/新加卷/pocamen/dataset/bulbasaur/00000000.png