deepcompute / cnn Goto Github PK

基于Java实现CNN，并附MNIST和语音（MFCC特征）性别识别示例。

Java 100.00%

cnn's Introduction

JavaCNN

基于Java实现CNN

构建CNN

LayerBuilder builder = new LayerBuilder();
builder.addLayer(Layer.buildInputLayer(new Size(28, 28)));
builder.addLayer(Layer.buildConvLayer(6, new Size(5, 5)));
builder.addLayer(Layer.buildSampLayer(new Size(2, 2)));
builder.addLayer(Layer.buildConvLayer(12, new Size(5, 5)));
builder.addLayer(Layer.buildSampLayer(new Size(2, 2)));
builder.addLayer(Layer.buildOutputLayer(10));
CNN cnn = new CNN(builder, 50);

运行MNIST数据集

String fileName = "data/train.format";
Dataset dataset = Dataset.load(fileName, ",", 784);
cnn.train(dataset, 100);
Dataset testset = Dataset.load("data/test.format", ",", -1);
cnn.predict(testset, "data/test.predict");

计算精度可以达到97.8%。

cnn's People

Contributors

Stargazers

Watchers

Forkers

brysj chunhub donganwangshi wangyx0055 weifenghu flyinggh arleyliu jingshengzhao gx-2016 walletma 764073542 ungar7 mobangjack wondery deyuwang chenxicui rohithkodali rongyousu jsonbao yulifei jaciyu dasonmo cloudstdio jfdw phantom201485015 cosastro zsh965866221 killf luckynote ooole leon-trainee ryfan-rs yiqinggit zjucsxxd pingchunzhang tangww901127 nysyxxg longweibing chentianson yszddai ocean-jiang wuhaiweigit leeqiang250 chenx376 dangyue sgccerpi sajian stevejamesyang hanlos midalas longsy316 jingchun01 rian-day mrpady renallen yangshuangjia hutiaotiao sh1neon bhtype0129 jiang-s feibabm nolenhuang luodebug awesom1-tech kevin-ke jiyulongxu fivecp mingmchen hongweizhang97 cc8848 907552421 jinxiu0406 twilightbi nicccccck china-liweihong onlysdy feng20068 wangbaochao xrh666 xiejiajun moonlight15111 hider87 snow-1314 zxyd518 shangqingb newcoding-0

cnn's Issues

can not find "mnist/model/model.cnn"

Exception in thread "main" java.lang.RuntimeException: java.io.FileNotFoundException: mnist/model/model.cnn (没有那个文件或目录)
    at info.hb.cnn.core.CNN.saveModel(CNN.java:571)
    at info.hb.ccn.main.CNNMnist.runTrain(CNNMnist.java:45)
    at info.hb.ccn.main.CNNMnist.main(CNNMnist.java:23)
    at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
    at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
    at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
    at java.lang.reflect.Method.invoke(Method.java:498)
    at com.intellij.rt.execution.application.AppMain.main(AppMain.java:147)
Caused by: java.io.FileNotFoundException: mnist/model/model.cnn (没有那个文件或目录)
    at java.io.FileOutputStream.open0(Native Method)
    at java.io.FileOutputStream.open(FileOutputStream.java:270)
    at java.io.FileOutputStream.<init>(FileOutputStream.java:213)
    at java.io.FileOutputStream.<init>(FileOutputStream.java:101)
    at info.hb.cnn.core.CNN.saveModel(CNN.java:566)
    ... 7 more

没有 mnist/model/model.cnn 这个文件哒，要怎么获取呢？

speech data？

请问speech示例中，输入文件每一行代表一帧的MFCC系数和还是一行代表一个音频的mfcc信息？从mfcc输出的结果信息需经过转换再输入cnn吗？求训练集与测试集文件格式介绍啊~

如何构建训练数据集

我如何自己构建训练数据呢，以及如何指定训练结果集，train.format 的结构是怎么样的

输出一堆乱码，请问改用哪个字符集呢

如题

A error while Size.x!=Size.y

In CNN.java,
Line 445:inputLayer.setMapValue(0, i, j, attr[mapSize.x * i + j]);
Will throw an error while Size.x!=Size.y, This line should be fixed as below:

inputLayer.setMapValue(0, i, j, attr[mapSize.y * i + j]);

用于提取图片特征的类是哪个类？

这个可以用来给图像分类吗

outmaps[m] * (1 - outmaps[m]) * (target[m] - outmaps[m]) 不太明白是啥意思

speech data format

可以说一下那个 train format的格式吗？就是一行中前36个数是mfcc后的系数还是什么？