學習Tensorflow或深度學習，難免用到各種資料集， 最近用到cifar10資料集，簡單研究了下，然後把cifar-10資料集儲存為jpg圖片，分別利用python和c++做了實現。

關於cifar-10，網上介紹很多，這裡主要用了python和binary版本：

python版 

    每個batch包含一個字典，該字典有data和labels兩個key。其中，data是1000*3072（ 3 *32 *32）的影象資料。1000即圖片數量，前1024個數據是red通道畫素值，然後1024是個green通道畫素值，最後啥blue通道。labels是1000個0~9表示資料類別的資料。

程式碼如下：

import numpy as np
from PIL import Image
import pickle
import os

CHANNEL = 3
WIDTH = 32
HEIGHT = 32

data = []
labels=[]
classification = ['airplane','automobile','bird','cat','deer','dog','frog','horse','ship','truck']

for i in range(5):
    with open("data/cifar-10-batches-py/data_batch_"+ str(i+1),mode='rb') as file:
        data_dict = pickle.load(file, encoding='bytes')
        data+= list(data_dict[b'data'])
        labels+= list(data_dict[b'labels'])

img =  np.reshape(data,[-1,CHANNEL, WIDTH, HEIGHT])


data_path = "data/images/"
if not os.path.exists(data_path):
    os.makedirs(data_path)
for i in range(img.shape[0]):

    r = img[i][0]
    g = img[i][1]
    b = img[i][2]

    ir = Image.fromarray(r)
    ig = Image.fromarray(g)
    ib = Image.fromarray(b)
    rgb = Image.merge("RGB", (ir, ig, ib))

    name = "img-" + str(i) +"-"+ classification[labels[i]]+ ".png"
    rgb.save(data_path + name, "PNG")



 

結果截圖：

C++版 

每個batch包括10000*（1 + 3072）大小資料，1代表label大小，3072是影象資料。儲存方式同上。

程式碼如下：

#include<iostream>
#include<opencv2/opencv.hpp>

using namespace std;
using namespace cv;

#define WIDTH 32
#define HEIGHT 32
#define CHANNEL 3
#define PERNUM 1000 
#define CLASS 10


char classification[CLASS][256] = { "airplane", "automobile", "bird", "cat", "deer", "dog", "frog", "horse", "ship", "truck" };

int main(){

	FILE *pBatch = fopen("data_batch_1.bin","rb");
	if (!pBatch)
		return -1;

	unsigned char buf[CHANNEL * WIDTH * HEIGHT + 1];
	memset(buf,0,sizeof(buf));
	Mat bgr;
	bgr.create(WIDTH,HEIGHT,CV_8UC3);

	int index = 0;
	while (!feof(pBatch)){
		
		fread(buf, 1, CHANNEL * WIDTH * HEIGHT + 1, pBatch);
		unsigned char* pBuf = buf + 1;

		for (int i = 0; i < bgr.rows;i++){

			Vec3b *pbgr = bgr.ptr<Vec3b>(i);
			
			for (int j = 0; j < bgr.cols;j++){
				//pBuf += (i * bgr.rows + j * bgr.cols);
				
				for (int c = 0; c < 3;c++){

					pbgr[j][c] = pBuf[(2 - c)* bgr.rows * bgr.cols + i * bgr.rows + j ];
				}

			}
		}
		
		imwrite("image/img" + to_string(index)+".jpg",bgr);
		index++;
	}

	fclose(pBatch);
	return 0;
}
 

結果截圖：