Mnist 图片测试

1、目的

在上述训练模型的基础上，测试自己手写的图片，完成手写图片的生成、图片转caffe输入格式文件生成以及模型测试。

2、手写图片、caffe输入格式转换

windows画图软件，属性设置图片像素28*28

图片黑底白字，

用matlab将图片转化为mnist二进制格式文件，

clc;
clear;
for i=1:10
imgOrigin=imread([num2str(i,'%d'),'.png']);
imgBinaryImg(i,:,:)= imgOrigin(:,:,1).'; 
end
imgMnist = fopen('test_img_ubyte', 'wb');  
magic = 2051;
numImages = i;
numRows = 28;
numCols = 28;  
fwrite(imgMnist, magic, 'int32', 0, 'ieee-be'); 
fwrite(imgMnist, numImages, 'int32', 0, 'ieee-be');
fwrite(imgMnist, numRows, 'int32', 0, 'ieee-be');  
fwrite(imgMnist, numCols, 'int32', 0, 'ieee-be');  
for i=1:10
imgBinaryImg1(numCols* numRows*(i-1)+1:numCols* numRows*i)= reshape(imgBinaryImg(i,:,:), 1, numCols* numRows); 
end
fwrite(imgMnist, imgBinaryImg1, 'unsigned char');  
fclose(imgMnist);  
flabel = fopen('test_lable_ubyte', 'wb');  
magic = 2049;
numLabels = i;
labels = [5 6 1 4 7 0 9 1 3 5];
fwrite(flabel, magic, 'int32', 0, 'ieee-be');  
fwrite(flabel, numLabels, 'int32', 0, 'ieee-be');  
fwrite(flabel, labels, 'unsigned char');  
fclose(flabel);  

运行后产生文件

将其放于\caffe\data\mnist文件夹。
git下运行/mnist/create_mnist_byourownselves.sh ，产生caffe读入文件

#!/usr/bin/env sh
# This script converts the mnist data into lmdb/leveldb format,
# depending on the value assigned to $BACKEND.
set -e
EXAMPLE=examples/mnist
DATA=data/mnist
BUILD=Build/x64/Release
BACKEND="lmdb"
echo "Creating ${BACKEND}..."
rm -rf $EXAMPLE/mnist_train_${BACKEND}
rm -rf $EXAMPLE/mnist_test_${BACKEND}
$BUILD/convert_mnist_data.exe $DATA/train-images-idx3-ubyte \
  $DATA/train-labels-idx1-ubyte $EXAMPLE/mnist_train_${BACKEND} --backend=${BACKEND}
$BUILD/convert_mnist_data.exe $DATA/test_img_ubyte \
  $DATA/test_lable_ubyte $EXAMPLE/mnist_test_${BACKEND} --backend=${BACKEND}
echo "Done."

产生测试图片为10张，修改lenet_train_test.prototxt

batch_size为10，
git bash下运行,

./Build/x64/Release/caffe.exe test -model examples/mnist/lenet_train_test_byourself.prototxt  -weights examples/mnist/lenet_iter_10000.caffemodel  -iterations 1

测试结果为：

所有代码以及资料地址：

https://github.com/Alix1992/Caffe

参考博客：

https://blog.csdn.net/woyaopojie1990/article/details/50820526`