判断caffe安装是否生效：运行caffe自带的mnist示例

你的名字 2022-01-28 21:19 265阅读 0赞

1、先进入caffe文件目录，再用data/mnist下的get\_mnist.sh下载MNIST数据集，代码如下：

sudo sh ./data/mnist/get\_mnist.sh

打开下载目录caffe/data/mnist查看如下图：

![20190520102029952.png][]

2、转换格式

./examples/mnist/create\_mnist.sh

![20190520102400460.png][]

![20190520102408811.png][]

数据内容：

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80MTc3MDE2OQ_size_16_color_FFFFFF_t_70][]

训练网络：

网络结构定义在./examples/mnist/lenet\_train\_test.prototxt中。

训练参数配置在./examples/mnist/lenet\_solver.prototxt中。

3、训练

通过命令行执行训练，代码如下：

cd ./caffe

./build/tools/caffe train --solver=examples/mnist/lenet\_solver.prototxt

gpu只用了20s的时间就训练好：

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80MTc3MDE2OQ_size_16_color_FFFFFF_t_70 1][]

正确率：99.04%

输出模型：

路径：/home/xuqiong/caffe/examples/mnist

![20190520103300386.png][]

4、测试

cd到caffe路径下：

./build/tools/caffe.bin test -model=examples/mnist/lenet\_train\_test.prototxt -weights=examples/mnist/lenet\_iter\_10000.caffemodel

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80MTc3MDE2OQ_size_16_color_FFFFFF_t_70 2][]

参考：

[https://www.cnblogs.com/wmlj/p/8681216.html][https_www.cnblogs.com_wmlj_p_8681216.html]

[https://blog.csdn.net/AlexQiweek/article/details/51281240][https_blog.csdn.net_AlexQiweek_article_details_51281240]

[20190520102029952.png]: /images/20220129/4902ea43756340d1a2faf78bf6e05094.png
[20190520102400460.png]: /images/20220129/665142dfc94248529a4c9aa01e846f69.png
[20190520102408811.png]: /images/20220129/3199f3f694ef4bbdabafb5741d0304cb.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80MTc3MDE2OQ_size_16_color_FFFFFF_t_70]: /images/20220129/b74561fe27254ef5bff081b339e33f7c.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80MTc3MDE2OQ_size_16_color_FFFFFF_t_70 1]: /images/20220129/63626d019de44c7f86036a70c9d333fd.png
[20190520103300386.png]: /images/20220129/d329a179baa448378f352aca04458281.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80MTc3MDE2OQ_size_16_color_FFFFFF_t_70 2]: /images/20220129/30a7bcf076ad4e28b049d040df1294a5.png
[https_www.cnblogs.com_wmlj_p_8681216.html]: https://www.cnblogs.com/wmlj/p/8681216.html
[https_blog.csdn.net_AlexQiweek_article_details_51281240]: https://blog.csdn.net/AlexQiweek/article/details/51281240