当前位置： 代码迷 >> 综合 >> 【DeepLearning】【Caffe】基于Caffe官方教程简单介绍caffe和pycaffe接口使用方法
  详细解决方案

【DeepLearning】【Caffe】基于Caffe官方教程简单介绍caffe和pycaffe接口使用方法

热度：6   发布时间：2023-12-06 13:27:04.0

本文基于Caffe官方教程中的手写体数字识别项目记录caffe和pycaffe的使用方法.

文章目录

      • 1. MNIST和LeNet
      • 2. 使用caffe命令行接口训练模型
        • 2.1 准备数据集
        • 2.2 定义 LeNet 模型
        • 2.3 训练 LeNet
      • 3. 使用pycaffe接口训练模型
        • 3.1 设置环境
        • 3.2 在 jupyter notebook 中工作

1. MNIST和LeNet

MNIST 是一个手写体数字识别数据集. 它是从 National Institute of Standards and Technology (NIST) 的庞大数据集中构建的一个子集.

LeNet 是 Lecun 设计的卷积神经网络，用于手写数字分类任务.

本文基于【DeepLearning】【Caffe】编译caffe及虚拟python环境的pycaffe接口 部署的环境.

2. 使用caffe命令行接口训练模型

Training LeNet on MNIST with Caffe

2.1 准备数据集

1、 进入caffe-master目录，$CAFFE_ROOT是caffe-master的绝对路径.

$ cd $CAFFE_ROOT

2、 下载并解压数据集. 在$CAFFE_ROOT/data/mnist/得到train-images-idx3-ubyte、train-labels-idx1-ubyte、t10k-images-idx3-ubyte、t10k-labels-idx1-ubyte.

$ ./data/mnist/get_mnist.sh

3、 将原始数据集转换为lmdb格式. 训练集在$CAFFE_ROOT/examples/mnist/mnist_train_lmdb目录下，测试集在$CAFFE_ROOT/examples/mnist/mnist_test_lmdb.

$ ./examples/mnist/create_mnist.sh
2.2 定义 LeNet 模型

caffe 命令行工具使用 Google Protobuf 定义模型结构和优化方法. 在 src/caffe/proto/caffe.proto 中可以查看 caffe 用到的设置.

1、 在 examples/mnist/lenet_train_test.prototxt 中定义 LeNet 的模型结构.

定义模型名称：

name: "LeNet"

定义数据层，从 lmdb 中读取数据：

layer {name: "mnist"type: "Data"transform_param {scale: 0.00390625}data_param {source: "mnist_train_lmdb"backend: LMDBbatch_size: 64}top: "data"top: "label"
}

该层name为 mnist，type为 data，从source读取数据，backend定义数据格式，batch_size大小为64. scale为 0.00390625 = 1 / 256 0.00390625=1/256 0.00390625=1/256，用于将像素值标准化到 [ 0 , 1 ) [0,1) [0,1)区间. top表示该层的输出，该层产生两个 blobs，一个是 data blob，另一个是 label blob.

定义卷积层：

layer {name: "conv1"type: "Convolution"param { lr_mult: 1 }param { lr_mult: 2 }convolution_param {num_output: 20kernel_size: 5stride: 1weight_filler {type: "xavier"}bias_filler {type: "constant"}}bottom: "data"top: "conv1"
}

该层接收 data blob，神经元数量为 num_output . 卷积核大小为 kernel_size，步长为 stride，weight_filler 和 bias_filler 分别定义初始权重参数和偏置参数. lr_mult 表示该层自适应的学习率，1 表示该层权重参数的学习率和优化器定义的学习率相同，2 表示该层偏置参数的学习率是优化器定义的学习率的2倍. bottom 表示该层的输入，top 表示该层的输出.

定义池化层：

layer {name: "pool1"type: "Pooling"pooling_param {kernel_size: 2stride: 2pool: MAX}bottom: "conv1"top: "pool1"
}

该层定义比较简单，kernel_size为池化核大小，stride为步长，pool定义池化方式.

定义全连接层：

layer {name: "ip1"type: "InnerProduct"param { lr_mult: 1 }param { lr_mult: 2 }inner_product_param {num_output: 500weight_filler {type: "xavier"}bias_filler {type: "constant"}}bottom: "pool2"top: "ip1"
}

该层定义也很简单，神经元数量为 num_output，其它与前面相似.

定义激活层：

layer {name: "relu1"type: "ReLU"bottom: "ip1"top: "ip1"
}

该层定义激活函数，type 为激活函数类型. 如果该层支持 in-place 操作，bottom 和 top 可以同名，节省存储空间.

定义损失层：

layer {name: "loss"type: "SoftmaxWithLoss"bottom: "ip2"bottom: "label"
}

最后该层定义损失函数，type为损失函数类型. 该层接收两个blobs计算损失值，没有输出. 当反向传播开始计算时，该层产生对上一层的梯度.

定义 Layer Rules ：
各层使用include定义该层在何时处于模型的结构中，例如：

layer {// ...layer definition...include: { phase: TRAIN }
}

在这个例子中，该层仅在模型训练阶段被包含在内. 如果将TRAIN改为TEST，则该层仅在测试阶段被包含在内. 默认的，该层没有include定义的 layer rules 时总是在网络模型之中. 因此，在 lenet_train_test.prototxt 中定义了两个 Data 层（它们的 batch_size 不同），一个用于训练阶段，另一个用于测试阶段. 同样的，Accuracy 层仅在测试阶段出现在模型之中.

2、 在 examples/mnist/lenet_solver.prototxt 中定义训练 LeNet 的优化方法.

# The train/test net protocol buffer definition
net: "examples/mnist/lenet_train_test.prototxt"
# test_iter specifies how many forward passes the test should carry out.
# In the case of MNIST, we have test batch size 100 and 100 test iterations,
# covering the full 10,000 testing images.
test_iter: 100
# Carry out testing every 500 training iterations.
test_interval: 500
# The base learning rate, momentum and the weight decay of the network.
base_lr: 0.01
momentum: 0.9
weight_decay: 0.0005
# The learning rate policy
lr_policy: "inv"
gamma: 0.0001
power: 0.75
# Display every 100 iterations
display: 100
# The maximum number of iterations
max_iter: 10000
# snapshot intermediate results
snapshot: 5000
snapshot_prefix: "examples/mnist/lenet"
# solver mode: CPU or GPU
solver_mode: GPU
2.3 训练 LeNet

执行下面脚本开始训练 LeNet

$ ./examples/mnist/train_lenet.sh

该脚本的执行的具体命令如下

#!/usr/bin/env sh
set -e./build/tools/caffe train --solver=examples/mnist/lenet_solver.prototxt $@

执行训练脚本后，命令行开始输出训练过程信息. 最开始为硬件信息.

I1107 20:52:36.823984 10432 caffe.cpp:204] Using GPUs 0
I1107 20:52:37.016600 10432 caffe.cpp:209] GPU 0: GeForce GTX 1070

从 examples/mnist/lenet_solver.prototxt 初始化优化器

I1107 20:52:38.889734 10432 solver.cpp:45] Initializing solver from parameters:

从 examples/mnist/lenet_train_test.prototxt 创建训练模型

I1107 20:52:38.925832 10432 solver.cpp:102] Creating training net from net file: examples/mnist/lenet_train_test.prototxt
I1107 20:52:38.926434 10432 net.cpp:296] The NetState phase (0) differed from the phase (1) specified by a rule in layer mnist
I1107 20:52:38.926487 10432 net.cpp:296] The NetState phase (0) differed from the phase (1) specified by a rule in layer accuracy
I1107 20:52:38.926750 10432 net.cpp:53] Initializing net from parameters:

开始逐层初始化，例如下面是创建 conv1 的输出信息. 这些信息对debug非常有用.

I1107 20:52:38.967473 10432 net.cpp:86] Creating Layer conv1
I1107 20:52:38.967494 10432 net.cpp:408] conv1 <- data
I1107 20:52:38.967540 10432 net.cpp:382] conv1 -> conv1
I1107 20:52:43.570137 10432 net.cpp:124] Setting up conv1
I1107 20:52:43.570161 10432 net.cpp:131] Top shape: 64 20 24 24 (737280)
I1107 20:52:43.570166 10432 net.cpp:139] Memory required for data: 3150080

训练模型初始化完成.

I1107 20:52:43.579761 10432 net.cpp:257] Network initialization done.

从 examples/mnist/lenet_train_test.prototxt 创建测试模型

I1107 20:52:43.579917 10432 solver.cpp:190] Creating test net (#0) specified by net file: examples/mnist/lenet_train_test.prototxt
I1107 20:52:43.579960 10432 net.cpp:296] The NetState phase (1) differed from the phase (0) specified by a rule in layer mnist
I1107 20:52:43.580045 10432 net.cpp:53] Initializing net from parameters:

测试模型也要初始化

I1107 20:52:43.640388 10432 net.cpp:86] Creating Layer conv1
I1107 20:52:43.640391 10432 net.cpp:408] conv1 <- data
I1107 20:52:43.640398 10432 net.cpp:382] conv1 -> conv1
I1107 20:52:43.642616 10432 net.cpp:124] Setting up conv1
I1107 20:52:43.642632 10432 net.cpp:131] Top shape: 100 20 24 24 (1152000)
I1107 20:52:43.642637 10432 net.cpp:139] Memory required for data: 4922800

测试模型初始化完成

I1107 20:52:43.649025 10432 net.cpp:257] Network initialization done.

开始训练过程

I1107 20:52:43.649063 10432 solver.cpp:57] Solver scaffolding done.
I1107 20:52:43.649327 10432 caffe.cpp:239] Starting Optimization
I1107 20:52:43.649333 10432 solver.cpp:289] Solving LeNet

在优化选项中，设置每隔100次迭代打印训练信息

I1107 20:52:44.042378 10432 solver.cpp:239] Iteration 100 (561.408 iter/s, 0.178124s/100 iters), loss = 0.210162
I1107 20:52:44.042407 10432 solver.cpp:258]     Train net output #0: loss = 0.210162 (* 1 = 0.210162 loss)
I1107 20:52:44.042412 10432 sgd_solver.cpp:112] Iteration 100, lr = 0.00992565

每隔500次迭代测试一次模型

I1107 20:52:44.659152 10432 solver.cpp:347] Iteration 500, Testing net (#0)
I1107 20:52:44.717660 10443 data_layer.cpp:73] Restarting data prefetching from start.
I1107 20:52:44.719583 10432 solver.cpp:414]     Test net output #0: accuracy = 0.9743
I1107 20:52:44.719601 10432 solver.cpp:414]     Test net output #1: loss = 0.083975 (* 1 = 0.083975 loss)

每隔5000次迭代保存一次模型参数和优化设置

I1107 20:52:52.069522 10432 solver.cpp:464] Snapshotting to binary proto file examples/mnist/lenet_iter_5000.caffemodel
I1107 20:52:52.153425 10432 sgd_solver.cpp:284] Snapshotting solver state to binary proto file examples/mnist/lenet_iter_5000.solverstate

训练完成后测试模型

I1107 20:53:00.412792 10432 solver.cpp:347] Iteration 10000, Testing net (#0)
I1107 20:53:00.473321 10443 data_layer.cpp:73] Restarting data prefetching from start.
I1107 20:53:00.474253 10432 solver.cpp:414]     Test net output #0: accuracy = 0.9911
I1107 20:53:00.474272 10432 solver.cpp:414]     Test net output #1: loss = 0.029594 (* 1 = 0.029594 loss)
I1107 20:53:00.474277 10432 solver.cpp:332] Optimization Done.
I1107 20:53:00.474280 10432 caffe.cpp:250] Optimization Done.

3. 使用pycaffe接口训练模型

Learning LeNet

01-learning-lenet.ipynb 位于 $CAFFE_ROOT/examples/ 目录下.

3.1 设置环境
$ workon caffe-master

(caffe-master)$ pip install ipykernel

(caffe-master)$ python -m ipykernel install --user --name caffe-master --display-name "caffe-master"

(caffe-master)$ cd $CAFFE_ROOT/examples/

(caffe-master)$ jupyter notebook
3.2 在 jupyter notebook 中工作

打开 jupyter notebook 后如下所示
在这里插入图片描述

新建 notebook，命名为 Solving in Python with LeNet
在这里插入图片描述

在这里插入图片描述
In [1]:

from pylab import *
%matplotlib inline

In [2]:

caffe_root = '../'  # this file should be run from {caffe_root}/examples (otherwise change this line)

在《3.3 配置pycaffe接口依赖》中已经设置过 PYTHONPATH，可以直接导入. 如果没有设置，使用 sys.path 方法导入.

In [3]:

import caffe

使用 python 定义 LeNet 模型.

In [4]:

from caffe import layers as L, params as Pdef lenet(lmdb, batch_size):# our version of LeNet: a series of linear and simple nonlinear transformationsn = caffe.NetSpec()n.data, n.label = L.Data(batch_size=batch_size, backend=P.Data.LMDB, source=lmdb,transform_param=dict(scale=1./255), ntop=2)n.conv1 = L.Convolution(n.data, kernel_size=5, num_output=20, weight_filler=dict(type='xavier'))n.pool1 = L.Pooling(n.conv1, kernel_size=2, stride=2, pool=P.Pooling.MAX)n.conv2 = L.Convolution(n.pool1, kernel_size=5, num_output=50, weight_filler=dict(type='xavier'))n.pool2 = L.Pooling(n.conv2, kernel_size=2, stride=2, pool=P.Pooling.MAX)n.fc1 =   L.InnerProduct(n.pool2, num_output=500, weight_filler=dict(type='xavier'))n.relu1 = L.ReLU(n.fc1, in_place=True)n.score = L.InnerProduct(n.relu1, num_output=10, weight_filler=dict(type='xavier'))n.loss =  L.SoftmaxWithLoss(n.score, n.label)return n.to_proto()

转换为 protobuf 以便 Caffe 读取.

In [5]:

with open('mnist/lenet_auto_train.prototxt', 'w') as f:f.write(str(lenet('mnist/mnist_train_lmdb', 64)))with open('mnist/lenet_auto_test.prototxt', 'w') as f:f.write(str(lenet('mnist/mnist_test_lmdb', 100)))

查看训练模型.

In [6]:

!cat mnist/lenet_auto_train.prototxt

layer {
    name: "data"type: "Data"top: "data"top: "label"transform_param {
    scale: 0.003921568859368563}data_param {
    source: "mnist/mnist_train_lmdb"batch_size: 64backend: LMDB}
}
layer {
    name: "conv1"type: "Convolution"bottom: "data"top: "conv1"convolution_param {
    num_output: 20kernel_size: 5weight_filler {
    type: "xavier"}}
}
layer {
    name: "pool1"type: "Pooling"bottom: "conv1"top: "pool1"pooling_param {
    pool: MAXkernel_size: 2stride: 2}
}
layer {
    name: "conv2"type: "Convolution"bottom: "pool1"top: "conv2"convolution_param {
    num_output: 50kernel_size: 5weight_filler {
    type: "xavier"}}
}
layer {
    name: "pool2"type: "Pooling"bottom: "conv2"top: "pool2"pooling_param {
    pool: MAXkernel_size: 2stride: 2}
}
layer {
    name: "fc1"type: "InnerProduct"bottom: "pool2"top: "fc1"inner_product_param {
    num_output: 500weight_filler {
    type: "xavier"}}
}
layer {
    name: "relu1"type: "ReLU"bottom: "fc1"top: "fc1"
}
layer {
    name: "score"type: "InnerProduct"bottom: "fc1"top: "score"inner_product_param {
    num_output: 10weight_filler {
    type: "xavier"}}
}
layer {
    name: "loss"type: "SoftmaxWithLoss"bottom: "score"bottom: "label"top: "loss"
}

查看测试模型.

In [7]:

!cat mnist/lenet_auto_test.prototxt

layer {
    name: "data"type: "Data"top: "data"top: "label"transform_param {
    scale: 0.003921568859368563}data_param {
    source: "mnist/mnist_test_lmdb"batch_size: 100backend: LMDB}
}
layer {
    name: "conv1"type: "Convolution"bottom: "data"top: "conv1"convolution_param {
    num_output: 20kernel_size: 5weight_filler {
    type: "xavier"}}
}
layer {
    name: "pool1"type: "Pooling"bottom: "conv1"top: "pool1"pooling_param {
    pool: MAXkernel_size: 2stride: 2}
}
layer {
    name: "conv2"type: "Convolution"bottom: "pool1"top: "conv2"convolution_param {
    num_output: 50kernel_size: 5weight_filler {
    type: "xavier"}}
}
layer {
    name: "pool2"type: "Pooling"bottom: "conv2"top: "pool2"pooling_param {
    pool: MAXkernel_size: 2stride: 2}
}
layer {
    name: "fc1"type: "InnerProduct"bottom: "pool2"top: "fc1"inner_product_param {
    num_output: 500weight_filler {
    type: "xavier"}}
}
layer {
    name: "relu1"type: "ReLU"bottom: "fc1"top: "fc1"
}
layer {
    name: "score"type: "InnerProduct"bottom: "fc1"top: "score"inner_product_param {
    num_output: 10weight_filler {
    type: "xavier"}}
}
layer {
    name: "loss"type: "SoftmaxWithLoss"bottom: "score"bottom: "label"top: "loss"
}
!cat mnist/lenet_auto_sol

查看优化器设置.

In [8]:

!cat mnist/lenet_auto_solver.prototxt

# The train/test net protocol buffer definition
train_net: "mnist/lenet_auto_train.prototxt"
test_net: "mnist/lenet_auto_test.prototxt"
# test_iter specifies how many forward passes the test should carry out.
# In the case of MNIST, we have test batch size 100 and 100 test iterations,
# covering the full 10,000 testing images.
test_iter: 100
# Carry out testing every 500 training iterations.
test_interval: 500
# The base learning rate, momentum and the weight decay of the network.
base_lr: 0.01
momentum: 0.9
weight_decay: 0.0005
# The learning rate policy
lr_policy: "inv"
gamma: 0.0001
power: 0.75
# Display every 100 iterations
display: 100
# The maximum number of iterations
max_iter: 10000
# snapshot intermediate results
snapshot: 5000
snapshot_prefix: "mnist/lenet"

选择硬件设备.

In [9]:

caffe.set_device(0)
caffe.set_mode_gpu()

选择优化算法并加载优化设置》

In [10]:

solver = None  # ignore this workaround for lmdb data (can't instantiate two solvers on the same data)
solver = caffe.SGDSolver('mnist/lenet_auto_solver.prototxt')

查看各层中间结果的张量维度.

In [11]:

# each output is (batch size, feature dim, spatial dim)
[(k, v.data.shape) for k, v in solver.net.blobs.items()]

Out[11]:

[('data', (64, 1, 28, 28)),('label', (64,)),('conv1', (64, 20, 24, 24)),('pool1', (64, 20, 12, 12)),('conv2', (64, 50, 8, 8)),('pool2', (64, 50, 4, 4)),('fc1', (64, 500)),('score', (64, 10)),('loss', ())]

查看权重变量的张量维度.

In [12]:

# just print the weight sizes (we'll omit the biases)
[(k, v[0].data.shape) for k, v in solver.net.params.items()]

Out[12]:

[('conv1', (20, 1, 5, 5)),('conv2', (50, 20, 5, 5)),('fc1', (500, 800)),('score', (10, 500))]

训练模型前向计算.

In [13]:

solver.net.forward()  # train net

Out[13]:

{
    'loss': array(2.3712316, dtype=float32)}

测试模型前向计算.

In [14]:

solver.test_nets[0].forward()  # test net (there can be more than one)

Out[14]:

{
    'loss': array(2.4383156, dtype=float32)}

查看前8个训练图像和标签.

In [15]:

# we use a little trick to tile the first eight images
imshow(solver.net.blobs['data'].data[:8, 0].transpose(1, 0, 2).reshape(28, 8*28), cmap='gray'); axis('off')
print('train labels:', solver.net.blobs['label'].data[:8])

train labels: [5. 0. 4. 1. 9. 2. 1. 3.]

在这里插入图片描述

查看前8个测试图像和标签.

In [16]:

imshow(solver.test_nets[0].blobs['data'].data[:8, 0].transpose(1, 0, 2).reshape(28, 8*28), cmap='gray'); axis('off')
print('test labels:', solver.test_nets[0].blobs['label'].data[:8])

test labels: [7. 2. 1. 0. 4. 1. 4. 9.]

在这里插入图片描述

计算一次反向传播.

In [17]:

solver.step(1)

查看第一层的梯度值. 4x5的网格中每一格都是5x5的卷积核.

In [18]:

imshow(solver.net.params['conv1'][0].diff[:, 0].reshape(4, 5, 5, 5).transpose(0, 2, 1, 3).reshape(4*5, 5*5), cmap='gray'); axis('off')

Out[18]:

(-0.5, 24.5, 19.5, -0.5)

在这里插入图片描述

自定义一个迭代过程.

In [19]:

%%time
niter = 200
test_interval = 25
# losses will also be stored in the log
train_loss = zeros(niter)
test_acc = zeros(int(np.ceil(niter / test_interval)))
output = zeros((niter, 8, 10))# the main solver loop
for it in range(niter):solver.step(1)  # SGD by Caffe# store the train losstrain_loss[it] = solver.net.blobs['loss'].data# store the output on the first test batch# (start the forward pass at conv1 to avoid loading new data)solver.test_nets[0].forward(start='conv1')output[it] = solver.test_nets[0].blobs['score'].data[:8]# run a full test every so often# (Caffe can also do this for us and write to a log, but we show here# how to do it directly in Python, where more complicated things are easier.)if it % test_interval == 0:print('Iteration', it, 'testing...')correct = 0for test_it in range(100):solver.test_nets[0].forward()correct += sum(solver.test_nets[0].blobs['score'].data.argmax(1)== solver.test_nets[0].blobs['label'].data)test_acc[it // test_interval] = correct / 1e4

Iteration 0 testing...
Iteration 25 testing...
Iteration 50 testing...
Iteration 75 testing...
Iteration 100 testing...
Iteration 125 testing...
Iteration 150 testing...
Iteration 175 testing...
CPU times: user 1.91 s, sys: 551 ms, total: 2.46 s
Wall time: 1.41 s

查看训练损失和测试精度关于迭代次数的曲线.

In [20]:

_, ax1 = subplots()
ax2 = ax1.twinx()
ax1.plot(arange(niter), train_loss)
ax2.plot(test_interval * arange(len(test_acc)), test_acc, 'r')
ax1.set_xlabel('iteration')
ax1.set_ylabel('train loss')
ax2.set_ylabel('test accuracy')
ax2.set_title('Test Accuracy: {:.2f}'.format(test_acc[-1]))

Out[20]:

Text(0.5, 1.0, 'Test Accuracy: 0.94')

在这里插入图片描述

查看前8个测试数据，和 LeNet 对它们的预测结果. 预测结果为长度为10的一维向量，分别表示 0~9 的置信度得分. 这里的预测结果是全连接层的直接输出，没有经过softmax激活函数. 第一横栏图像是第 1 个测试图片，手写数字 7. 第二横栏图像是 LeNet 迭代过程中 50 次测试结果. 横坐标表示使用第几次测试时的模型，它反映了迭代次数. 纵坐标表示预测标签 0~9. 像素的亮度表示置信度得分，越亮得分越高. 可以看出随着迭代次数增加，标签为 7 的那一行像素越来越亮. 在最后的标签向量中，7 是最亮的.

In [21]:

for i in range(8):figure(figsize=(2, 2))imshow(solver.test_nets[0].blobs['data'].data[i, 0], cmap='gray')figure(figsize=(10, 2))imshow(output[:50, i].T, interpolation='nearest', cmap='gray')xlabel('iteration')ylabel('label')

在这里插入图片描述
在这里插入图片描述
在这里插入图片描述
在这里插入图片描述
在这里插入图片描述
在这里插入图片描述
在这里插入图片描述
在这里插入图片描述
在这里插入图片描述
在这里插入图片描述
在这里插入图片描述
在这里插入图片描述
在这里插入图片描述
在这里插入图片描述
在这里插入图片描述
在这里插入图片描述

从上往下这 8 个测试图片越来越困难. 第八张测试图片“9”看起来很像“4”. LeNet 对其预测的结果中，4 和 9 都有很高的置信度得分.

下面展示了 softmax 激活后的结果. 可以看出，预测结果更为明显了.

In [22]:

for i in range(8):figure(figsize=(2, 2))imshow(solver.test_nets[0].blobs['data'].data[i, 0], cmap='gray')figure(figsize=(10, 2))imshow(exp(output[:50, i].T) / exp(output[:50, i].T).sum(0), interpolation='nearest', cmap='gray')xlabel('iteration')ylabel('label')

在这里插入图片描述
在这里插入图片描述
在这里插入图片描述
在这里插入图片描述
在这里插入图片描述
在这里插入图片描述
在这里插入图片描述
在这里插入图片描述
在这里插入图片描述
在这里插入图片描述
在这里插入图片描述
在这里插入图片描述
在这里插入图片描述
在这里插入图片描述
在这里插入图片描述
在这里插入图片描述

自定义修改网络结构.

In [23]:

train_net_path = 'mnist/custom_auto_train.prototxt'
test_net_path = 'mnist/custom_auto_test.prototxt'
solver_config_path = 'mnist/custom_auto_solver.prototxt'### define net
def custom_net(lmdb, batch_size):# define your own net!n = caffe.NetSpec()# keep this data layer for all networksn.data, n.label = L.Data(batch_size=batch_size, backend=P.Data.LMDB, source=lmdb,transform_param=dict(scale=1./255), ntop=2)# EDIT HERE to try different networks# this single layer defines a simple linear classifier# (in particular this defines a multiway logistic regression)n.score =   L.InnerProduct(n.data, num_output=10, weight_filler=dict(type='xavier'))# EDIT HERE this is the LeNet variant we have already tried# n.conv1 = L.Convolution(n.data, kernel_size=5, num_output=20, weight_filler=dict(type='xavier'))# n.pool1 = L.Pooling(n.conv1, kernel_size=2, stride=2, pool=P.Pooling.MAX)# n.conv2 = L.Convolution(n.pool1, kernel_size=5, num_output=50, weight_filler=dict(type='xavier'))# n.pool2 = L.Pooling(n.conv2, kernel_size=2, stride=2, pool=P.Pooling.MAX)# n.fc1 = L.InnerProduct(n.pool2, num_output=500, weight_filler=dict(type='xavier'))# EDIT HERE consider L.ELU or L.Sigmoid for the nonlinearity# n.relu1 = L.ReLU(n.fc1, in_place=True)# n.score = L.InnerProduct(n.fc1, num_output=10, weight_filler=dict(type='xavier'))# keep this loss layer for all networksn.loss =  L.SoftmaxWithLoss(n.score, n.label)return n.to_proto()with open(train_net_path, 'w') as f:f.write(str(custom_net('mnist/mnist_train_lmdb', 64)))    
with open(test_net_path, 'w') as f:f.write(str(custom_net('mnist/mnist_test_lmdb', 100)))### define solver
from caffe.proto import caffe_pb2
s = caffe_pb2.SolverParameter()# Set a seed for reproducible experiments:
# this controls for randomization in training.
s.random_seed = 0xCAFFE# Specify locations of the train and (maybe) test networks.
s.train_net = train_net_path
s.test_net.append(test_net_path)
s.test_interval = 500  # Test after every 500 training iterations.
s.test_iter.append(100) # Test on 100 batches each time we test.s.max_iter = 10000     # no. of times to update the net (training iterations)# EDIT HERE to try different solvers
# solver types include "SGD", "Adam", and "Nesterov" among others.
s.type = "SGD"# Set the initial learning rate for SGD.
s.base_lr = 0.01  # EDIT HERE to try different learning rates
# Set momentum to accelerate learning by
# taking weighted average of current and previous updates.
s.momentum = 0.9
# Set weight decay to regularize and prevent overfitting
s.weight_decay = 5e-4# Set `lr_policy` to define how the learning rate changes during training.
# This is the same policy as our default LeNet.
s.lr_policy = 'inv'
s.gamma = 0.0001
s.power = 0.75
# EDIT HERE to try the fixed rate (and compare with adaptive solvers)
# `fixed` is the simplest policy that keeps the learning rate constant.
# s.lr_policy = 'fixed'# Display the current training loss and accuracy every 1000 iterations.
s.display = 1000# Snapshots are files used to store networks we've trained.
# We'll snapshot every 5K iterations -- twice during training.
s.snapshot = 5000
s.snapshot_prefix = 'mnist/custom_net'# Train on the GPU
s.solver_mode = caffe_pb2.SolverParameter.GPU# Write the solver to a temporary file and return its filename.
with open(solver_config_path, 'w') as f:f.write(str(s))### load the solver and create train and test nets
solver = None  # ignore this workaround for lmdb data (can't instantiate two solvers on the same data)
solver = caffe.get_solver(solver_config_path)### solve
niter = 250  # EDIT HERE increase to train for longer
test_interval = niter / 10
# losses will also be stored in the log
train_loss = zeros(niter)
test_acc = zeros(int(np.ceil(niter / test_interval)))# the main solver loop
for it in range(niter):solver.step(1)  # SGD by Caffe# store the train losstrain_loss[it] = solver.net.blobs['loss'].data# run a full test every so often# (Caffe can also do this for us and write to a log, but we show here# how to do it directly in Python, where more complicated things are easier.)if it % test_interval == 0:print('Iteration', it, 'testing...')correct = 0for test_it in range(100):solver.test_nets[0].forward()correct += sum(solver.test_nets[0].blobs['score'].data.argmax(1)== solver.test_nets[0].blobs['label'].data)test_acc[int(it // test_interval)] = correct / 1e4_, ax1 = subplots()
ax2 = ax1.twinx()
ax1.plot(arange(niter), train_loss)
ax2.plot(test_interval * arange(len(test_acc)), test_acc, 'r')
ax1.set_xlabel('iteration')
ax1.set_ylabel('train loss')
ax2.set_ylabel('test accuracy')
ax2.set_title('Custom Test Accuracy: {:.2f}'.format(test_acc[-1]))

Iteration 0 testing...
Iteration 25 testing...
Iteration 50 testing...
Iteration 75 testing...
Iteration 100 testing...
Iteration 125 testing...
Iteration 150 testing...
Iteration 175 testing...
Iteration 200 testing...
Iteration 225 testing...

Out[23]:

Text(0.5, 1.0, 'Custom Test Accuracy: 0.88')

在这里插入图片描述

  相关解决方案

代码迷 - 您身边的软件异常，代码异常，编程异常助手（发现，解决，分享）
Copyright © 2009-2013 DAIMAMI.COM. All rights reserved.