deeplabv2实现并训练自己的数据

一、安装必要的依赖库 matio

安装方法1：

sudo apt-get install libmatio-dev

安装方法2：

下载matio(https://sourceforge.net/projects/matio/files/matio/1.5.2/)

tar zxf matio-1.5.2.tar.gz

cd matio-1.5.2

./configure

make

make check

make install

sudo ldconfig

二、下载Deeplabv2并编译

deeplab_v2官方源码地址如下：https://bitbucket.org/aquariusjay/deeplab-public-ver2/overview

1.下载deeplab源码进行编译。

mkdir deeplab

cd deeplab

git clone https://bitbucket.org/aquariusjay/deeplab-public-ver2.git

然后进行编译，其实deeplab-public-ver2就是caffe,所以编译方法和caffe相同

make all -j16

make pycaffe

make test

如何存在问题可以参看 https://blog.csdn.net/ziyouyi111/article/details/79358401和 https://blog.csdn.net/ziyouyi111/article/details/80274528

2.依次建立存放设置文件夹，预测结果文件夹，数据集txt文件夹，log文件夹，model文件夹，evaluation文件夹，并将

官方给的txt文件夹，以及预训练的model和网络结构文件放到相应的文件夹中。这里省略比较繁琐的步骤使用 https://github.com/xmojiao/deeplab_v2.git 中voc12文件夹即可。

3.deeplab2的script文件run_pascal.sh 解析

#!/bin/sh

## MODIFY PATH for YOUR SETTING

ROOT_DIR= #数据路径设计为空载train.txt和test.txt文件中需要写成全路径

CAFFE_DIR=../code #设置为deeplab-public-ver2的caffe路径

CAFFE_BIN=${CAFFE_DIR}/.build_release/tools/caffe.bin

EXP=. #设为当前路径

if [ "${EXP}" = "." ]; then

NUM_LABELS=2 #分割类别一定要修改

DATA_ROOT=${ROOT_DIR} #VOC数据目录，修改为你的数据目录

else

NUM_LABELS=0

echo "Wrong exp name"

## Specify which model to train

########### voc12 ################

NET_ID=deelab_largeFOV ##此处文件名有问题应该改为deeplab_largeFOV

## Variables used for weakly or semi-supervisedly training

#TRAIN_SET_SUFFIX=

#TRAIN_SET_SUFFIX=_aug #此处应该取消注释，当你run training 1时

#TRAIN_SET_STRONG=train

#TRAIN_SET_STRONG=train200

#TRAIN_SET_STRONG=train500

#TRAIN_SET_STRONG=train1000

#TRAIN_SET_STRONG=train750

#TRAIN_SET_WEAK_LEN=5000

DEV_ID=0

#####

## Create dirs

CONFIG_DIR=${EXP}/config/${NET_ID} #此处目录为/voc12/config/deeplab_largeFOV

MODEL_DIR=${EXP}/model/${NET_ID}

mkdir -p ${MODEL_DIR} #创建MODEL_DIR目录为/voc12/model/deeplab_largeFOV

LOG_DIR=${EXP}/log/${NET_ID}

mkdir -p ${LOG_DIR}

export GLOG_log_dir=${LOG_DIR}

## Run

RUN_TRAIN=1 #为1 0说明执行train

RUN_TEST=0 #为0 1说明执行test

RUN_TRAIN2=0

RUN_TEST2=0

## Training #1 (on train_aug)

if [ ${RUN_TRAIN} -eq 1 ]; then #r如果RUN_TRAIN为1

LIST_DIR=${EXP}/list

TRAIN_SET=train${TRAIN_SET_SUFFIX}

if [ -z ${TRAIN_SET_WEAK_LEN} ]; then #如果TRAIN_SET_WEAK_LEN长度为零则为真

TRAIN_SET_WEAK=${TRAIN_SET}_diff_${TRAIN_SET_STRONG}

comm -3 ${LIST_DIR}/${TRAIN_SET}.txt ${LIST_DIR}/${TRAIN_SET_STRONG}.txt > ${LIST_DIR}/${TRAIN_SET_WEAK}.txt #comm -3 指令为不输出两个文件共有的行，此处即为除去train.txt文件中train_aug.txt的数据，其他都输出到train_aud_diff_train.txt

else

TRAIN_SET_WEAK=${TRAIN_SET}_diff_${TRAIN_SET_STRONG}_head${TRAIN_SET_WEAK_LEN}

comm -3 ${LIST_DIR}/${TRAIN_SET}.txt ${LIST_DIR}/${TRAIN_SET_STRONG}.txt | head -n ${TRAIN_SET_WEAK_LEN} > ${LIST_DIR}/${TRAIN_SET_WEAK}.txt

MODEL=${EXP}/model/${NET_ID}/init.caffemodel #下载的vgg16或者ResNet101中的 model

echo Training net ${EXP}/${NET_ID}

for pname in train solver; do

sed "$(eval echo $(cat sub.sed))" \

${CONFIG_DIR}/${pname}.prototxt > ${CONFIG_DIR}/${pname}_${TRAIN_SET}.prototxt #复制文件train.prototxt到train_train_train_aug.prototxt,slove同理

done #此部分运行时如以下命令

CMD="${CAFFE_BIN} train \

--solver=${CONFIG_DIR}/solver_${TRAIN_SET}.prototxt \

--gpu=${DEV_ID}"

if [ -f ${MODEL} ]; then

CMD="${CMD} --weights=${MODEL}"

echo Running ${CMD} && ${CMD}

#train部分运行时，即以下运行命令 ../deeplab-public-ver2/.build_release/tools/caffe.bin train --solver=volab_largeFOV/solver_train_aug.prototxt --gpu=0 --weights=voc12/model/deeplab_largeFOV/init.caf femodel

#上述命令中，solver_train_aug.prototxt由solve.prototxt文件复制而来，init.caffemodel为原始下载了的VGG16的model

## Test #1 specification (on val or test)

if [ ${RUN_TEST} -eq 1 ]; then

for TEST_SET in val; do

TEST_ITER=`cat ${EXP}/list/${TEST_SET}.txt | wc -l` #此处计算val.txt文件中测试图片个数，共1449个

MODEL=${EXP}/model/${NET_ID}/test.caffemodel

if [ ! -f ${MODEL} ]; then

MODEL=`ls -t ${EXP}/model/${NET_ID}/train_iter_*.caffemodel | head -n 1`

echo Testing net ${EXP}/${NET_ID}

FEATURE_DIR=${EXP}/features/${NET_ID}

mkdir -p ${FEATURE_DIR}/${TEST_SET}/fc8

mkdir -p ${FEATURE_DIR}/${TEST_SET}/fc9

mkdir -p ${FEATURE_DIR}/${TEST_SET}/seg_score

sed "$(eval echo $(cat sub.sed))" \

${CONFIG_DIR}/test.prototxt > ${CONFIG_DIR}/test_${TEST_SET}.prototxt

CMD="${CAFFE_BIN} test \

--model=${CONFIG_DIR}/test_${TEST_SET}.prototxt \

--weights=${MODEL} \

--gpu=${DEV_ID} \

--iterations=${TEST_ITER}"

echo Running ${CMD} && ${CMD}

done

#test部分运行时，即以下运行命令../deeplab-public-ver2/.build_release/tools/caffe.bin test --model=voc12/config/deeplab_largeFOV/test_val.prototxt --weights=voc12/model/deeplab_largeFOV/train_iter_20000.caffemodel --gpu=0 --iterations=1449

#上述命令中，test_val.prototxt由test.prototxt文件复制而来，train_iter_20000.caffemode由第一部分train得到的model

## Training #2 (finetune on trainval_aug)

if [ ${RUN_TRAIN2} -eq 1 ]; then

LIST_DIR=${EXP}/list

TRAIN_SET=trainval${TRAIN_SET_SUFFIX}

if [ -z ${TRAIN_SET_WEAK_LEN} ]; then

TRAIN_SET_WEAK=${TRAIN_SET}_diff_${TRAIN_SET_STRONG}

comm -3 ${LIST_DIR}/${TRAIN_SET}.txt ${LIST_DIR}/${TRAIN_SET_STRONG}.txt > ${LIST_DIR}/${TRAIN_SET_WEAK}.txt

else

TRAIN_SET_WEAK=${TRAIN_SET}_diff_${TRAIN_SET_STRONG}_head${TRAIN_SET_WEAK_LEN}

comm -3 ${LIST_DIR}/${TRAIN_SET}.txt ${LIST_DIR}/${TRAIN_SET_STRONG}.txt | head -n ${TRAIN_SET_WEAK_LEN} > ${LIST_DIR}/${TRAIN_SET_WEAK}.txt

MODEL=${EXP}/model/${NET_ID}/init2.caffemodel

if [ ! -f ${MODEL} ]; then

MODEL=`ls -t ${EXP}/model/${NET_ID}/train_iter_*.caffemodel | head -n 1`

echo Training2 net ${EXP}/${NET_ID}

for pname in train solver2; do

sed "$(eval echo $(cat sub.sed))" \

${CONFIG_DIR}/${pname}.prototxt > ${CONFIG_DIR}/${pname}_${TRAIN_SET}.prototxt

done

CMD="${CAFFE_BIN} train \

--solver=${CONFIG_DIR}/solver2_${TRAIN_SET}.prototxt \

--weights=${MODEL} \

--gpu=${DEV_ID}"

echo Running ${CMD} && ${CMD}

## Test #2 on official test set

if [ ${RUN_TEST2} -eq 1 ]; then

for TEST_SET in val test; do

TEST_ITER=`cat ${EXP}/list/${TEST_SET}.txt | wc -l`

MODEL=${EXP}/model/${NET_ID}/test2.caffemodel

if [ ! -f ${MODEL} ]; then

MODEL=`ls -t ${EXP}/model/${NET_ID}/train2_iter_*.caffemodel | head -n 1`

echo Testing2 net ${EXP}/${NET_ID}

FEATURE_DIR=${EXP}/features2/${NET_ID}

mkdir -p ${FEATURE_DIR}/${TEST_SET}/fc8

mkdir -p ${FEATURE_DIR}/${TEST_SET}/crf

sed "$(eval echo $(cat sub.sed))" \

${CONFIG_DIR}/test.prototxt > ${CONFIG_DIR}/test_${TEST_SET}.prototxt

CMD="${CAFFE_BIN} test \

--model=${CONFIG_DIR}/test_${TEST_SET}.prototxt \

--weights=${MODEL} \

--gpu=${DEV_ID} \

--iterations=${TEST_ITER}"

echo Running ${CMD} && ${CMD}

done

4.训练数据train.txt文件中的图像和标签，label图盘为二值图，如下图所示。

5.deeplab网络训练

将run_pascal.sh中的设置为run模式：

RUN_TRAIN=1

RUN_TEST=0

运行程序

cd voc2012

sh run_pascal.sh 2>&1|tee train.log

6.deeplab网络测试

RUN_TRAIN=0 #为1 0说明执行train

RUN_TEST=1 #为0 1说明执行test

sh run_pascal.sh 2>&1|tee test.log

7.mat转png图片

python create_labels_21.py 注意修改好路径

参考文献：

https://blog.csdn.net/ruotianxia/article/details/78331964

https://github.com/xmojiao/deeplab_v2

https://blog.csdn.net/Xmo_jiao/article/details/77897109?locationNum=11