紧接之前Yann Lecun的LeNet的论文Gradient-Based Learning Applied to Document Recognition,我们这边来实践一下LeNet网络
这里我们使用Caffe框架进行LeNet的实践,用的数据集是mnist
首先先下载mnist数据集;
在Caffe目录下执行:
./data/mnist/get_mnist.sh
./exampels/mnist/create_mnist.sh
获取mnist数据集并将数据集转换为mnist_train_lmdb和mnist_test_lmdb格式。
在caffe/example/mnist目录下有一个lenet_train_test.prototxt
name: "LeNet"
layer {
name: "mnist"//名字
type: "Data"//类型为数据
top: "data"
top: "label"
include {
phase: TRAIN
}
transform_param {
scale: 0.00390625//归一化 1/256
}
data_param {
source: "examples/mnist/mnist_train_lmdb"
batch_size: 64//分批处理的图像个数
backend: LMDB
}
}
layer {
name: "mnist"
type: "Data"
top: "data"//生成two blobs,分别为data blob 和label blob
top: "label"
include {
phase: TEST
}
transform_param {
scale: 0.00390625
}
data_param {
source: "examples/mnist/mnist_test_lmdb"
batch_size: 100
backend: LMDB
}
}
layer {
name: "conv1"
type: "Convolution"//卷积层
bottom: "data"
top: "conv1"
param {
lr_mult: 1//权重的学习率与solver运行的学习率一致
}
param {
lr_mult: 2//偏置的学习率是solver运行的学习率的2倍
}
convolution_param {
num_output: 20//20个featuremap
kernel_size: 5//卷积核为5x5
stride: 1//步长为1
weight_filler {
type: "xavier"//用 xavier算法初始化权重
}
bias_filler {
type: "constant"//用常数初始化权重
}
}
}
layer {
name: "pool1"
type: "Pooling"
bottom: "conv1"
top: "pool1"
pooling_param {
pool: MAX
kernel_size: 2//核大小2
stride: 2//步长大小2
}
}
layer {
name: "conv2"
type: "Convolution"
bottom: "pool1"
top: "conv2"
param {
lr_mult: 1
}
param {
lr_mult: 2
}
convolution_param {
num_output: 50
kernel_size: 5
stride: 1
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
}
}
}
layer {
name: "pool2"
type: "Pooling"
bottom: "conv2"
top: "pool2"
pooling_param {
pool: MAX
kernel_size: 2
stride: 2
}
}
layer {
name: "ip1"
type: "InnerProduct"
bottom: "pool2"
top: "ip1"
param {
lr_mult: 1
}
param {
lr_mult: 2
}
inner_product_param {
num_output: 500
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
}
}
}
layer {
name: "relu1"
type: "ReLU"
bottom: "ip1"
top: "ip1"
}
layer {
name: "ip2"
type: "InnerProduct"
bottom: "ip1"
top: "ip2"
param {
lr_mult: 1
}
param {
lr_mult: 2
}
inner_product_param {
num_output: 10
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
}
}
}
layer {
name: "accuracy"
type: "Accuracy"
bottom: "ip2"
bottom: "label"
top: "accuracy"
include {
phase: TEST
}
}
layer {
name: "loss"
type: "SoftmaxWithLoss"
bottom: "ip2"
bottom: "label"
top: "loss"
}
用Netscope可视化的结果:
(这边用relu代替了原来的sigmoid函数)
在/caffe/examples/mnist/lenet_solver.prototxt中有训练参数的配置:
# The train/test net protocol buffer definition//使bi用网络结构
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//测试时bitch_size100
# Carry out testing every 500 training iterations.//每500轮测试一次
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每100次迭代次数显示一次训练时lr和loss
display: 100
# The maximum number of iterations最大迭代次数
max_iter: 10000
# snapshot intermediate results每5000次迭代输出模型
snapshot: 5000
snapshot_prefix: "examples/mnist/lenet"//模型保存路径
# solver mode: CPU or GPU
solver_mode: GPU
通过运行
./examples/mnist/train_lenet.sh
train_lenet.sh:
#!/usr/bin/env sh
set -e
./build/tools/caffe train --solver=examples/mnist/lenet_solver.prototxt $@
可以进行训练,下面是训练结果:
可以看到准确率可以达到99.03%,已经是相当高了,但网络对于像mnist效果好,但对于稍大一点的数据集效果就会下降地很明显。
可以看到caffe能快速地进行网络的搭建,但是相对于tensorflow来说不够灵活,而且caffe无法进行递归循环网络的搭建。