快捷導(dǎo)航

使用darknet框架的imagenet數(shù)據(jù)分類預(yù)訓(xùn)練操作

更新時間：2020年07月07日 10:35:22 作者：baidu_38371925

這篇文章主要介紹了使用darknet框架的imagenet數(shù)據(jù)分類預(yù)訓(xùn)練操作，具有很好的參考價值，希望對大家有所幫助。一起跟隨小編過來看看吧

最近一段時間一直在研究yolo物體檢測，基于網(wǎng)絡(luò)上很少有yolo的分類預(yù)訓(xùn)練和yolo9000的聯(lián)合數(shù)據(jù)的訓(xùn)練方法，經(jīng)過本人的真實(shí)實(shí)驗，對這兩個部分做一個整理（本篇介紹yolo的分類預(yù)訓(xùn)練）

1、數(shù)據(jù)準(zhǔn)備

1000類的Imagenet圖片數(shù)據(jù)

因為Imagenet不同的類別數(shù)據(jù)都是單獨(dú)放在一個文件夾中，并且有特定的命名，如‘n00020287'，所以在做分類時我們不需要去制作特定的標(biāo)簽，只要訓(xùn)練的圖片的path中包含自身的類別標(biāo)簽，而不含有其他類的標(biāo)簽即可。

制作用于訓(xùn)練的數(shù)據(jù)列表*classf_list.txt

2、分類標(biāo)簽制作

制作所有類別的標(biāo)簽列表new_label.txt和標(biāo)簽對應(yīng)的類別名稱的列表new_name.txt

new_label.txt

new_name.txt（訓(xùn)練時不需要，但是測試時可以顯示出具體的類別）

3、修改cfg/.data配置文件（*classf.data）

classes=1000
train =/home/research/disk2/wangshun/yolo1700/darknet/coco/filelist/classf_list.txt
labels=data/new_label.txt
names=data/new_name.txt
backup=backup
top=5

修改網(wǎng)絡(luò)配置文件(classf.cfg)

[net]
#Training
batch=64
subdivisions=1
width=416
height=416
channels=3
momentum=0.9
decay=0.0005
angle=0
saturation = 1.5
exposure = 1.5
hue=.1
max_crop = 512
learning_rate=0.001
burn_in=1000
max_batches = 1000000000
policy=steps
steps=350000,500000,750000,1200000
scales=.1,.1,.1,.1
[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
[convolutional]
batch_normalize=1
filters=32
size=1
stride=1
pad=1
activation=leaky
[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
[convolutional]
batch_normalize=1
filters=64
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=128
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
[convolutional]
batch_normalize=1
filters=128
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=256
size=3
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=128
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=256
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
[convolutional]
batch_normalize=1
filters=128
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=256
size=3
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=128
size=1
stride=1
pad=1
activation=leaky
[convolutional]
batch_normalize=1
filters=256
size=3
stride=1
pad=1
activation=leaky
#######
[convolutional]
batch_normalize=1
size=1
stride=1
pad=1
filters=128
activation=leaky
[convolutional]
batch_normalize=1
size=3
stride=1
pad=1
filters=256
activation=leaky
[convolutional]
batch_normalize=1
size=3
stride=1
pad=1
filters=256
activation=leaky
[convolutional]
filters=1000
size=1
stride=1
pad=1
activation=leaky
[avgpool]
[softmax]
groups = 1
[cost]
type=sse

當(dāng)然中間的網(wǎng)絡(luò)層是我自己修改的網(wǎng)絡(luò)。

5. 訓(xùn)練