参考博客:blog.csdn.net/drrlalala/article/details/47274549
1,首先在网上下载图片,猫和狗。
直接保存下载该网页,会生成一个有图片的文件夹。
caffe-master/data 新建 myself
myself/ 新建 train dog
cat
test dog
cat
之后将图片分别复制到文件夹中
2,生成train.txt和test.txt
在train文件夹下运行 find -name *.jpg |cut -d '/' -f2-3 >train.txt
手动加标签 sed -i 's/$/ 0/g' train.txt 在末位加0
val.txt同理可得,也要加标签。
3,图片转化为lmdb格式
caffe-master/examples/ 新建 myself
复制examples/imagenet/create_imagenet.sh到 examples/myself
修改参照下面,主要是针对路径的修改
#!/usr/bin/env sh # Create the imagenet lmdb inputs # N.B. set the path to the imagenet train + val data dirs set -e EXAMPLE=examples/myself ##路径需要自己修改,默认的相对路径是在caffe-master下 DATA=data/myself ##是指生成的train.txt和val.txt的路径 TOOLS=build/tools TRAIN_DATA_ROOT=data/myself/train/ ##注生成的数据最前面就不需要加/了 VAL_DATA_ROOT=data/myself/test/ # Set RESIZE=true to resize the images to 256x256. Leave as false if images have # already been resized using another tool. RESIZE=true ##如果填true说明事先没有将其转为256*256格式 if $RESIZE; then RESIZE_HEIGHT=256 RESIZE_WIDTH=256 else RESIZE_HEIGHT=0 RESIZE_WIDTH=0 fi if [ ! -d "$TRAIN_DATA_ROOT" ]; then echo "Error: TRAIN_DATA_ROOT is not a path to a directory: $TRAIN_DATA_ROOT" echo "Set the TRAIN_DATA_ROOT variable in create_imagenet.sh to the path" "where the ImageNet training data is stored." exit 1 fi if [ ! -d "$VAL_DATA_ROOT" ]; then echo "Error: VAL_DATA_ROOT is not a path to a directory: $VAL_DATA_ROOT" echo "Set the VAL_DATA_ROOT variable in create_imagenet.sh to the path" "where the ImageNet validation data is stored." exit 1 fi echo "Creating train lmdb..." GLOG_logtostderr=1 $TOOLS/convert_imageset --resize_height=$RESIZE_HEIGHT --resize_width=$RESIZE_WIDTH --shuffle $TRAIN_DATA_ROOT $DATA/train.txt ##之前生成的list $EXAMPLE/myself_train_lmdb ##修改名字 echo "Creating val lmdb..." GLOG_logtostderr=1 $TOOLS/convert_imageset --resize_height=$RESIZE_HEIGHT --resize_width=$RESIZE_WIDTH --shuffle $VAL_DATA_ROOT $DATA/val.txt $EXAMPLE/myself_test_lmdb echo "Done."
之后 cd caffe-master
运行 sudo ./examples/myself/create_imagenet.sh
在examples/myself/ 下生成两个lmdb文件夹
4,计算图像均值
复制examples/imagenet/make_imagenet_mean.sh到 examples/myself
使用make_imagenet_mean.sh计算图像均值,在data/myself 下产生imagenet_mean.binaryproto文件
相对路径仍为 caffe-master下,按照自己的文件修改路径。
之后 sudo ./examples/myself/make_imagenet_mean.sh
#!/usr/bin/env sh # Compute the mean image from the imagenet training lmdb # N.B. this is available in data/ilsvrc12 EXAMPLE=examples/myself DATA=data/myself TOOLS=build/tools $TOOLS/compute_image_mean $EXAMPLE/myself_train_lmdb $DATA/imagenet_mean.binaryproto echo "Done."
5,定义网络
主要是修改下面三个文件:
my_train.prototxt
my_test.prototxt
solver.prototxt
这三个文件可以从caffe的已有例子中复制过来。接下来需要自己修改,修改的地方主要是,一些文件的路径和输出层的个数(即类别数目)。
6,写训练脚本my_train_lenet.sh 如下:(我这个是根据mnist中的文件改过来的)
#!/usr/bin/env sh set -e ./build/tools/caffe train --solver=examples/myself/solver.prototxt $@
未完待续。