Windows下使用caffe进行VGG人脸识别深度神经网络模型的微调训练.docx

Windows下采用caffe进行VGG人脸识别深度神经网络模型的微调训练2016-8-29 本文介绍在Windows环境下，采用采用caffe进行VGG人脸识别深度神经网络模型的fine-tune(微调训练)。运行环境为：windows7+vs2013+matlab2015a+caffe(微软版)。1. 代码和文件准备代码caffe：/BVLC/caffe/tree/windows，这个只要采用VS2013编译好就可以用了，配置编译按照网页上的说明进行即可完成。vgg-face模型：http:/www.robots.ox.ac.uk/vgg/software/vgg_face/。该网页包括caffe，matconvnet，torch三个版本，下载caffe版本即可。2. 数据准备收集人脸图像数据及对应的标签数据，vgg处理的图片的尺寸是224*224，因此不符合尺寸要求的要对尺寸进行修正。然后采用caffe中的convert_imageset工具将人脸图像及对应的标签转换为leveldb格式的库文件。并采用caffe中的compute_image_mean工具计算leveldb库文件中所包含人脸的均值文件face_mean.binaryproto。将训练数据和评估数据的库文件分别放置在下述位置：J:/caffe-windows/vggface_mycmd/vggface_train_leveldbJ:/caffe-windows/vggface_mycmd/vggface_val_leveldb3. 训练网络第三部分就开始训练网络了，首先需要到vggface的官网上下载vggface的caffe模型（官网还包括matconvnet模型，试过finetuning太慢了，torch没有试过），下载好了，就会有两个文件，一个是VGG_FACE_totxt，一个是VGG_FACE.caffemodel(深度网络的模型文件)。先建立一个.prototxt文件，命名为vggface_train_totx (深度网络的模型框架定义文件)即可，把之前VGG_FACE_totxt的所有的复制过来，然后加入数据层。VGG_FACE.caffemodel(深度网络的模型文件)与vggface_train_totx (深度网络的模型框架定义文件)要相配套。(1). vggface_train_totxtname: VGG_ILSVRC_16_layers layers name: data type: DATA include phase: TRAIN transform_param crop_size: 224 mean_value: 104 mean_value: 117 mean_value: 123 mirror: true data_param source: J:/caffe-windows/vggface_mycmd/vggface_train_leveldb batch_size: 3 backend: LEVELDB top: data top: labellayers name: data type: DATA include phase: TEST transform_param crop_size: 224 mean_value: 104 mean_value: 117 mean_value: 123 mirror: false data_param source: vggface_mycmd/vggface_val_leveldb batch_size: 3 backend: LEVELDB top: data top: labellayers bottom: data top: conv1_1 name: conv1_1 type: CONVOLUTION convolution_param num_output: 64 pad: 1 kernel_size: 3 layers bottom: conv1_1 top: conv1_1 name: relu1_1 type: RELUlayers bottom: conv1_1 top: conv1_2 name: conv1_2 type: CONVOLUTION convolution_param num_output: 64 pad: 1 kernel_size: 3 layers bottom: conv1_2 top: conv1_2 name: relu1_2 type: RELUlayers bottom: conv1_2 top: pool1 name: pool1 type: POOLING pooling_param pool: MAX kernel_size: 2 stride: 2 layers bottom: pool1 top: conv2_1 name: conv2_1 type: CONVOLUTION convolution_param num_output: 128 pad: 1 kernel_size: 3 layers bottom: conv2_1 top: conv2_1 name: relu2_1 type: RELUlayers bottom: conv2_1 top: conv2_2 name: conv2_2 type: CONVOLUTION convolution_param num_output: 128 pad: 1 kernel_size: 3 layers bottom: conv2_2 top: conv2_2 name: relu2_2 type: RELUlayers bottom: conv2_2 top: pool2 name: pool2 type: POOLING pooling_param pool: MAX kernel_size: 2 stride: 2 layers bottom: pool2 top: conv3_1 name: conv3_1 type: CONVOLUTION convolution_param num_output: 256 pad: 1 kernel_size: 3 layers bottom: conv3_1 top: conv3_1 name: relu3_1 type: RELUlayers bottom: conv3_1 top: conv3_2 name: conv3_2 type: CONVOLUTION convolution_param num_output: 256 pad: 1 kernel_size: 3 layers bottom: conv3_2 top: conv3_2 name: relu3_2 type: RELUlayers bottom: conv3_2 top: conv3_3 name: conv3_3 type: CONVOLUTION convolution_param num_output: 256 pad: 1 kernel_size: 3 layers bottom: conv3_3 top: conv3_3 name: relu3_3 type: RELUlayers bottom: conv3_3 top: pool3 name: pool3 type: POOLING pooling_param pool: MAX kernel_size: 2 stride: 2 layers bottom: pool3 top: conv4_1 name: conv4_1 type: CONVOLUTION convolution_param num_output: 512 pad: 1 kernel_size: 3 layers bottom: conv4_1 top: conv4_1 name: relu4_1 type: RELUlayers bottom: conv4_1 top: conv4_2 name: conv4_2 type: CONVOLUTION convolution_param num_output: 512 pad: 1 kernel_size: 3 layers bottom: conv4_2 top: conv4_2 name: relu4_2 type: RELUlayers bottom: conv4_2 top: conv4_3 name: conv4_3 type: CONVOLUTION convolution_param num_output: 512 pad: 1 kernel_size: 3 layers bottom: conv4_3 top: conv4_3 name: relu4_3 type: RELUlayers bottom: conv4_3 top: pool4 name: pool4 type: POOLING pooling_param pool: MAX kernel_size: 2 stride: 2 layers bottom: pool4 top: conv5_1 name: conv5_1 type: CONVOLUTION convolution_param num_output: 512 pad: 1 kernel_size: 3 layers bottom: conv5_1 top: conv5_1 name: relu5_1 type: RELUlayers bottom: conv5_1 top: conv5_2 name: conv5_2 type: CONVOLUTION convolution_param num_output: 512 pad: 1 kernel_size: 3 layers bottom: conv5_2 top: conv5_2 name: relu5_2 type: RELUlayers bottom: conv5_2 top: conv5_3 name: conv5_3 type: CONVOLUTION convolution_param num_output: 512 pad: 1 kernel_size: 3 layers bottom: conv5_3 top: conv5_3 name: relu5_3 type: RELUlayers bottom: conv5_3 top: pool5 name: pool5 type: POOLING pooling_param pool: MAX kernel_size: 2 stride: 2 layers bottom: pool5 top: fc6 name: fc6 type: INNER_PRODUCT inner_product_param num_output: 4096 layers bottom: fc6 top: fc6 name: relu6 type: RELUlayers bottom: fc6 top: fc6 name: drop6 type: DROPOUT dropout_param dropout_ratio: 0.5 layers bottom: fc6 top: fc7 name: fc7 type: INNER_PRODUCT inner_product_param num_output: 4096 layers bottom: fc7 top: fc7 name: relu7 type: RELUlayers bottom: fc7 top: fc7 name: drop7 type: DROPOUT dropout_param dropout_ratio: 0.5 layers bottom: fc7 top: fnc8 name: fnc8 type: INNER_PRODUCT inner_product_param num_output: 6 layers bottom: fnc8 top: prob name: prob type: SOFTMAX所做的修改主要为在原文件的第一个layer之前添加数据层。拉到文件的最底部，有个num_output: 2622的那一层，这一层主要是概率输出层，就是softmax分类器层。因为vgg训练这个网络，用了2622个人，所以就是2622，现在可以根据自己的人的个数来设置，我用了6个人，把num_output: 2622改为了num_output: 6，并且把name：所有的fc8改为了fnc8，即对fc8改名，否则会报错。好了，现在网络搭建好。另外，还需修改以下内容：(2). vgg_totxt完成vggface_train_totxt模型框架定义文件后，还需要编制一个对神经网络进行微调训练的参数文件VGG_totxt。VGG_totxt代码如下：net: vggface_mycmd/vggface_