作者：wjmishuai
出处：/wjmishuai/article/details/50854162
一：下载训练集和测试集
训练集的下载地址:http://www.cbsr./english/CASIA-WebFace-Database.html
测试集的下载地址:http://vis-www.cs.umass.edu/lfw/
二:利用linux脚本生成训练集
#!/bin/bash#-----------------------------------------------------------------------------------------#生成训练集和验证集#made by 郭开#-----------------------------------------------------------------------------------------#保存图片的路径 PATH=/media/gk/9ec75485-26b1-471f-9b7b-d18554ca3fdd/aaecho "start..."#遍历文件夹for name in $PATH/webface_img/*;dovar=0#遍历文件夹中的图片for file_name in $name/*;do var=$(($var+1)); str=$file_name #保存图片路径str=${str#*img} #保存图片lable lable=${str#*/} lable=${lable%%/*} if [ "$var" = "1" ] || [ "$var" = "3" ] || [ "$var" = "5" ]; then #验证集 echo $str" "$lable>>/home/gk/val.txt else #测试集 echo $str" "$lable>>/home/gk/train.txt fi done done echo "Done."
生成的文件如下图:
三：将图片转换成LMDB格式
#!/usr/bin/env sh #----------------------------------------------------------------------------------------- # 将图片批量生成lmbd格式的数据存储#made by 郭开 #----------------------------------------------------------------------------------------- #保存生成的lmdb的目录 EXAMPLE=/media/gk/9ec75485-26b1-471f-9b7b-d18554ca3fdd/aa/lmdb_webface#train.txt和val.txt所在的目录DATA=/home/gk#转换图片的工具所在的目录 TOOLS=/home/gk/caffe-master/build/tools #图片所在的目录TRAIN_DATA_ROOT=/media/gk/9ec75485-26b1-471f-9b7b-d18554ca3fdd/aa/webface_imgVAL_DATA_ROOT=/media/gk/9ec75485-26b1-471f-9b7b-d18554ca3fdd/aa/webface_img# 设置 RESIZE=true 可以把图片resize成想要的尺寸。RESIZE=trueif $RESIZE; thenRESIZE_HEIGHT=227RESIZE_WIDTH=227elseRESIZE_HEIGHT=0RESIZE_WIDTH=0fiif [ ! -d "$TRAIN_DATA_ROOT" ]; thenecho "Error: TRAIN_DATA_ROOT is not a path to a directory: $TRAIN_DATA_ROOT"echo "Set the TRAIN_DATA_ROOT variable in create_imagenet.sh to the path" \"where the ImageNet training data is stored."exit 1fiif [ ! -d "$VAL_DATA_ROOT" ]; thenecho "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 1fiecho "Creating train lmdb..."GLOG_logtostderr=1 $TOOLS/convert_imageset \--resize_height=$RESIZE_HEIGHT \--resize_width=$RESIZE_WIDTH \--shuffle \$TRAIN_DATA_ROOT \$DATA/train.txt \$EXAMPLE/train_lmdbecho "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/val_lmdbecho "Done."
四：计算图片的mean或者设置scale：0.00390625
#!/usr/bin/env sh #----------------------------------------------------------------------------------------- # 计算图像均值#made by 郭开 #----------------------------------------------------------------------------------------- #lmdb格式的文件所在的路径EXAMPLE=/media/gk/9ec75485-26b1-471f-9b7b-d18554ca3fdd/aa/lmdb_webface#均值文件保存的路径DATA=/media/gk/44CA719BCA718A46 #转换图片的工具所在的目录TOOLS=/home/gk/caffe-master/build/tools $TOOLS/compute_image_mean $EXAMPLE/train_lmdb \ $DATA/mean.binaryproto echo "Done."
五：训练模型
./build/tools/caffe train --solver=models/bvlc_reference_caffenet/solver.prototxt
slover.prototxt
net: "/home/gk/caffe-master/examples/vgg/train.prototxt"test_iter: 100test_interval: 1000base_lr: 0.001lr_policy: "step"gamma: 0.95stepsize: 100000momentum: 0.9weight_decay: 0.0005display: 100max_iter: 5000000snapshot: 50000snapshot_prefix: "/home/gk/caffe-master/examples/DeepID/snapshot"solver_mode: GPUdevice_id:0#debug_info: true
tarin.prototxt
name: "VGG_FACE_16_layers"layer {top: "data_1"top: "label_1"name: "data_1"type: "Data"data_param {source: "/media/gk/9ec75485-26b1-471f-9b7b-d18554ca3fdd/aa/webface_lmdb/train"backend:LMDBbatch_size: 128}transform_param {mean_file: "/media/gk/9ec75485-26b1-471f-9b7b-d18554ca3fdd/aa/webface_lmdb/mean.binaryproto"mirror: true}include: { phase: TRAIN }}layer {top: "data_1"top: "label_1"name: "data_1"type: "Data"data_param {source: "/media/gk/9ec75485-26b1-471f-9b7b-d18554ca3fdd/aa/webface_lmdb/val"backend:LMDBbatch_size: 128}transform_param {mean_file: "/media/gk/9ec75485-26b1-471f-9b7b-d18554ca3fdd/aa/webface_lmdb/mean.binaryproto"mirror: true}include: {phase: TEST}}layers {bottom: "data"top: "conv1_1"name: "conv1_1"type: CONVOLUTIONconvolution_param {num_output: 64pad: 1kernel_size: 3}}layers {bottom: "conv1_1"top: "conv1_1"name: "relu1_1"type: RELU}layers {bottom: "conv1_1"top: "conv1_2"name: "conv1_2"type: CONVOLUTIONconvolution_param {num_output: 64pad: 1kernel_size: 3}}layers {bottom: "conv1_2"top: "conv1_2"name: "relu1_2"type: RELU}layers {bottom: "conv1_2"top: "pool1"name: "pool1"type: POOLINGpooling_param {pool: MAXkernel_size: 2stride: 2}}layers {bottom: "pool1"top: "conv2_1"name: "conv2_1"type: CONVOLUTIONconvolution_param {num_output: 128pad: 1kernel_size: 3}}layers {bottom: "conv2_1"top: "conv2_1"name: "relu2_1"type: RELU}layers {bottom: "conv2_1"top: "conv2_2"name: "conv2_2"type: CONVOLUTIONconvolution_param {num_output: 128pad: 1kernel_size: 3}}layers {bottom: "conv2_2"top: "conv2_2"name: "relu2_2"type: RELU}layers {bottom: "conv2_2"top: "pool2"name: "pool2"type: POOLINGpooling_param {pool: MAXkernel_size: 2stride: 2}}layers {bottom: "pool2"top: "conv3_1"name: "conv3_1"type: CONVOLUTIONconvolution_param {num_output: 256pad: 1kernel_size: 3}}layers {bottom: "conv3_1"top: "conv3_1"name: "relu3_1"type: RELU}layers {bottom: "conv3_1"top: "conv3_2"name: "conv3_2"type: CONVOLUTIONconvolution_param {num_output: 256pad: 1kernel_size: 3}}layers {bottom: "conv3_2"top: "conv3_2"name: "relu3_2"type: RELU}layers {bottom: "conv3_2"top: "conv3_3"name: "conv3_3"type: CONVOLUTIONconvolution_param {num_output: 256pad: 1kernel_size: 3}}layers {bottom: "conv3_3"top: "conv3_3"name: "relu3_3"type: RELU}layers {bottom: "conv3_3"top: "pool3"name: "pool3"type: POOLINGpooling_param {pool: MAXkernel_size: 2stride: 2}}layers {bottom: "pool3"top: "conv4_1"name: "conv4_1"type: CONVOLUTIONconvolution_param {num_output: 512pad: 1kernel_size: 3}}layers {bottom: "conv4_1"top: "conv4_1"name: "relu4_1"type: RELU}layers {bottom: "conv4_1"top: "conv4_2"name: "conv4_2"type: CONVOLUTIONconvolution_param {num_output: 512pad: 1kernel_size: 3}}layers {bottom: "conv4_2"top: "conv4_2"name: "relu4_2"type: RELU}layers {bottom: "conv4_2"top: "conv4_3"name: "conv4_3"type: CONVOLUTIONconvolution_param {num_output: 512pad: 1kernel_size: 3}}layers {bottom: "conv4_3"top: "conv4_3"name: "relu4_3"type: RELU}layers {bottom: "conv4_3"top: "pool4"name: "pool4"type: POOLINGpooling_param {pool: MAXkernel_size: 2stride: 2}}layers {bottom: "pool4"top: "conv5_1"name: "conv5_1"type: CONVOLUTIONconvolution_param {num_output: 512pad: 1kernel_size: 3}}layers {bottom: "conv5_1"top: "conv5_1"name: "relu5_1"type: RELU}layers {bottom: "conv5_1"top: "conv5_2"name: "conv5_2"type: CONVOLUTIONconvolution_param {num_output: 512pad: 1kernel_size: 3}}layers {bottom: "conv5_2"top: "conv5_2"name: "relu5_2"type: RELU}layers {bottom: "conv5_2"top: "conv5_3"name: "conv5_3"type: CONVOLUTIONconvolution_param {num_output: 512pad: 1kernel_size: 3}}
