Question

我有一组图像（1000张图像）。每个人的维度为3072。每个图像都有这样的表示：例如图像1：

array([255, 78, 48, ..., 190, 230, 178], dtype=uint8)

我希望将其存储在矩阵中，使得每条线代表图像的矢量（3072）。这意味着我得到一个矩阵（1000,3072）这是我试过的

matrix_of_images= []
for img in images:
    data.append(img)
    data.append(img2)

但是附加列表给了我一个难以操作的结构，因为我想将它存储在csv文件中然后调用图像的一部分。

[array([255, 255, 255, ..., 255, 255, 255], dtype=uint8), array([0, 0, 0, ..., 0, 0, 0], dtype=uint8), array([0, 0, 0, ..., 0, 0, 0], dtype=uint8), array([255, 255, 255, ..., 255, 255, 255], dtype=uint8), array([255, 255, 255, ...,   0,   0,   0], dtype=uint8), array([  0,   0,   0, ..., 255, 255, 255], dtype=uint8), array([255, 255, 255, ...,   0,   0,   0], dtype=uint8), array([  0,   0,   0, ..., 255, 255, 255], dtype=uint8), array([255, 255, 255, ...,   0,   0,   0], dtype=uint8), array([  0,   0,   0, ..., 255, 255, 255], dtype=uint8)]

我正在寻找类似的东西 X =

[

[23,56, 78,....,45,156],
[60,56, 104,....,145,157],
[78,45, 7,....,0,15],
[45,56, 178,....,5,200]

]

其中l可以读取表单示例

X[1] #  second image
 [60,56, 104,....,145,157]
X[1][2] # third pixel of second image
104

一种易于存储在csv文件中的结构，其中列中的每个像素。

编辑： 在每次迭代时添加的向量是img1和img2

for i in range(1,500):

    #get coordinates
    #coords=npa[i,:]
    coords=npa.iloc[[i]]
    charac=characs[i-1]
    if (charac== "'/'"):
        charac= "'slash'"

    charac = charac.strip('"\'')
    #actual cropping of the image (easy with numpy)
    #img_charac=img[int(coords[2]):int(coords[4]),int(coords[3]):int(coords[5])]
    img_charac = img[int(coords[4]):int(coords[5]), int(coords[2]):int(coords[3])]
    #cv2.imwrite(path_save_cropped_images + str(charac) + "_" + str(i) + "_" + str(img_charac.shape) + ".png",  img_charac)

    #resize
    img_charac_resized=cv2.resize(img_charac, (32, 32), interpolation=cv2.INTER_NEAREST)
    #cv2.imwrite(path_save_resized_images + str(charac) + "_" + str(i) + "_" + str(img_charac_resized.shape) + ".png",img_charac_resized)
    #img_charac = cv2.resize(img_charac, (32, 32))

    #switch images
    img_charac_switched = 255 - img_charac_resized
    #cv2.imwrite(path_save_switched_pixels+ str(charac) +"_switched"+ "_" + str(i) + "_" + str(img_charac_switched.shape) + ".png",img_charac_switched)

    img1 = img_charac_resized.reshape((-1, 1))
    img1 = img1.T
    img1= img1.flatten()
    img1_label=charac

    img2=img_charac_switched.reshape((-1,1))
    img2=img2.T
    img2=img2.flatten()
    img2_label = charac
    #x=switch(charac)
    #saving the image

    #dataset

    #cv2.imwrite(path_dataset+ str(charac) + "_switched" + "_" + str(i) + "_" + str(img_charac_switched.shape) + ".png",img_charac_switched)
    #cv2.imwrite(path_dataset + str(charac) + "_" + str(i) + "_" + str(img_charac_resized.shape) + ".png",  img_charac_resized)


    #images = [img1,img2]
    img_arr = np.stack(img1, axis=0)
    img_arr = np.stack(img2, axis=0)
    #data.append(img1)
    #data.append(img2)

    #print (img_arr.shape)
    #print(i)



print(img_arr)
print(img_arr.shape)

Answer 1

您想要使用numpy.stack。指定axis=0垂直堆叠：

import numpy as np
n = 1000
images = [np.random.random(3072) for _ in range(n)]
img_arr = np.stack(images,axis=0)

>>> img_arr.shape
(1000, 3072)

如果是您的代码：

images = []
for i in range(500):
    # create img1 and img2
    images.extend([img1,img2])
img_arr = np.stack(images,axis=0)

用于存储和操作高效像素的数据结构：

2 个答案: