基于pycaffe从零开始写mnist（第二篇）——生成训练网络结构文件（train.prototxt）+测试网络结构文件（test.prototxt）

没什么好说的，直接贴代码，主要参考：http://blog.csdn.net/c406495762/article/details/70306795

# -*- coding: utf-8 -*-

__author__ = 'xuy'


# -*- coding: UTF-8 -*-
import caffe                                                     #导入caffe包
from datetime import datetime

def create_net(img_list, batch_size, include_acc=False):#用来创建train_val.prototxt,最终创建的是alexNet的网络结构
    #网络规范
    net = caffe.NetSpec()
    #ImageData数据层
    net.data, net.labels = caffe.layers.ImageData(batch_size = batch_size,#用来设置数据源的格式，由于数据源的格式是txt文件，因此类型是ImageData
                                                source = img_list,#数据源路径
                                                transform_param = dict(scale = 1./255),
                                                ntop = 2)
    #卷积层
    net.conv1 = caffe.layers.Convolution(net.data, kernel_size = 5, num_output = 20,
                                         weight_filler = dict(type = 'xavier'))
    #池化层
    net.pool1 = caffe.layers.Pooling(net.conv1, kernel_size = 2, stride = 2,
                                     pool = caffe.params.Pooling.MAX)
    #卷积层
    net.conv2 = caffe.layers.Convolution(net.pool1, kernel_size = 5, num_output = 50,
                                         weight_filler = dict(type = 'xavier'))
    #池化层
    net.pool2 = caffe.layers.Pooling(net.conv2, kernel_size = 2, stride = 2,
                                     pool = caffe.params.Pooling.MAX)
    #全连层
    net.fc1 =   caffe.layers.InnerProduct(net.pool2, num_output = 500,
                                          weight_filler = dict(type = 'xavier'))
    #激活函数层
    net.relu1 = caffe.layers.ReLU(net.fc1, in_place = True)
    #全连层
    net.score = caffe.layers.InnerProduct(net.relu1, num_output = 10,
                                          weight_filler = dict(type = 'xavier'))
    #softmax层
    net.loss =  caffe.layers.SoftmaxWithLoss(net.score, net.labels)

    if include_acc:
        net.acc = caffe.layers.Accuracy(net.score, net.labels)
        return net.to_proto()

    return net.to_proto()

def write_net(train_proto, train_list, test_proto, test_list):
    #写入prototxt文件
    with open(train_proto, 'w') as f:
        f.write(str(create_net(train_list, batch_size = 64)))
    #写入prototxt文件
    with open(test_proto, 'w') as f:
        f.write(str(create_net(test_list, batch_size = 100, include_acc = True)))

def write_sovler(my_project_root, solver_proto, train_proto, test_proto):#用来存储solver.prototxt文件
    sovler_string = caffe.proto.caffe_pb2.SolverParameter()                    #sovler存储
    sovler_string.train_net = train_proto                                    #train.prototxt位置指定
    sovler_string.test_net.append(test_proto)                                 #test.prototxt位置指定
    sovler_string.test_iter.append(100)                                        #10000/100 测试迭代次数
    sovler_string.test_interval = 938                                        #60000/64 每训练迭代test_interval次进行一次测试
    sovler_string.base_lr = 0.01                                            #基础学习率
    sovler_string.momentum = 0.9                                            #动量
    sovler_string.weight_decay = 5e-4                                        #权重衰减
    sovler_string.lr_policy = 'step'                                        #学习策略
    sovler_string.stepsize = 3000                                             #学习率变化频率
    sovler_string.gamma = 0.1                                                  #学习率变化指数
    sovler_string.display = 20                                                #每迭代display次显示结果
    sovler_string.max_iter = 9380                                            #10 epoch 938*10 最大迭代数
    sovler_string.snapshot = 938                                             #保存临时模型的迭代数
    sovler_string.snapshot_prefix = my_project_root + 'mnist'                #模型前缀
    sovler_string.solver_mode = caffe.proto.caffe_pb2.SolverParameter.GPU    #优化模式

    with open(solver_proto, 'w') as f:
        f.write(str(sovler_string))


def train(solver_proto):
    caffe.set_device(0)
    caffe.set_mode_gpu()
    solver = caffe.SGDSolver(solver_proto)#开始利用SGD方式开始训练
    solver.solve()


if __name__ == '__main__':
    my_project_root = "/home/xuy/桌面/code/python/caffe/python_mnist/"    #my-caffe-project目录
    train_list = my_project_root + "mnist/train/shuf_mytrain.txt"                #train.txt文件的位置
    test_list = my_project_root + "mnist/test/shuf_mytest.txt"                    #test.txt文件的位置
    train_proto = my_project_root + "mnist/train.prototxt"                #保存train.prototxt文件的位置,这个需要写入并且创建文件
    test_proto = my_project_root + "mnist/test.prototxt"                #保存test.prototxt文件的位置
    solver_proto = my_project_root + "mnist/solver.prototxt"            #保存solver.prototxt文件的位置

    write_net(train_proto, train_list, test_proto, test_list)
    print "生成train.prototxt test.prototxt成功"
    write_sovler(my_project_root, solver_proto, train_proto, test_proto)
    print "生成solver.prototxt成功"
    start_train_time=datetime.now()
    train(solver_proto)
    end_train_time=datetime.now()
    print "训练完成,训练时间是:%s"%(end_train_time-start_train_time).seconds