Question

我正在解决二进制分割问题。在测试之后，我使用Dice系数来计算测试掩模（地面实况）和相应的预测掩模之间的分割精度。我发现，在我的硬盘中保存之后，预测的掩码不是二进制的，而是大小为256X256X3。事实也是如此，即大小为256X256X3。在布尔转换之后，似乎所有像素值都被转换为true，用于预测的蒙版（第二个图像，img2）。为什么这样？这段代码是否正确计算了它们之间的Dice co-eff的值？

    from PIL import Image
    import numpy as np
    import matplotlib.pyplot as plt
    import cv2
    import matplotlib.pyplot as plt
    import matplotlib.image as mpimg

        def dice_coef(img, img2):
            if img.shape != img2.shape:
                raise ValueError("Shape mismatch: img and img2 must have to be of the same shape.")
            else:
                intersection = np.logical_and(img, img2)
                value = (2. * intersection.sum())  / (img.sum() + img2.sum())
            return value 

        img = cv2.imread('1.png')
        print(img)
        img = np.asarray(img).astype(np.bool)
        print(img)
        print(img.shape)
        print('_'*40)
        img2 = cv2.imread('2.png')
        print(img2)
        img2 = np.asarray(img2).astype(np.bool)
        print(img2)
        print(img2.shape)
        print('%'*40)
        value = dice_coef(img, img2)
        print(value)

输出：

[[[  0   0   0]
  [  0   0   0]
  [  0   0   0]
  ..., 
  [  0   0   0]
  [  0   0   0]
  [191 191 191]]

 [[  0   0   0]
  [  0   0   0]
  [  0   0   0]
  ..., 
  [  0   0   0]
  [  0   0   0]
  [255 255 255]]

 [[  0   0   0]
  [  0   0   0]
  [  0   0   0]
  ..., 
  [  0   0   0]
  [  0   0   0]
  [255 255 255]]

 ..., 
 [[  0   0   0]
  [  0   0   0]
  [  0   0   0]
  ..., 
  [157 157 157]
  [255 255 255]
  [255 255 255]]

 [[  0   0   0]
  [  0   0   0]
  [  0   0   0]
  ..., 
  [  0   0   0]
  [255 255 255]
  [255 255 255]]

 [[  0   0   0]
  [  0   0   0]
  [  0   0   0]
  ..., 
  [  0   0   0]
  [255 255 255]
  [255 255 255]]]
[[[False False False]
  [False False False]
  [False False False]
  ..., 
  [False False False]
  [False False False]
  [ True  True  True]]

 [[False False False]
  [False False False]
  [False False False]
  ..., 
  [False False False]
  [False False False]
  [ True  True  True]]

 [[False False False]
  [False False False]
  [False False False]
  ..., 
  [False False False]
  [False False False]
  [ True  True  True]]

 ..., 
 [[False False False]
  [False False False]
  [False False False]
  ..., 
  [ True  True  True]
  [ True  True  True]
  [ True  True  True]]

 [[False False False]
  [False False False]
  [False False False]
  ..., 
  [False False False]
  [ True  True  True]
  [ True  True  True]]

 [[False False False]
  [False False False]
  [False False False]
  ..., 
  [False False False]
  [ True  True  True]
  [ True  True  True]]]
(256, 256, 3)
________________________________________
[[[227 227 227]
  [243 243 243]
  [247 247 247]
  ..., 
  [  2   2   2]
  [  5   5   5]
  [  8   8   8]]

 [[233 233 233]
  [255 255 255]
  [250 250 250]
  ..., 
  [  2   2   2]
  [  4   4   4]
  [  6   6   6]]

 [[206 206 206]
  [252 252 252]
  [253 253 253]
  ..., 
  [  1   1   1]
  [  2   2   2]
  [  4   4   4]]

 ..., 
 [[  5   5   5]
  [  4   4   4]
  [  2   2   2]
  ..., 
  [214 214 214]
  [254 254 254]
  [230 230 230]]

 [[  6   6   6]
  [  5   5   5]
  [  4   4   4]
  ..., 
  [171 171 171]
  [207 207 207]
  [196 196 196]]

 [[  8   8   8]
  [  7   7   7]
  [  5   5   5]
  ..., 
  [ 81  81  81]
  [ 94  94  94]
  [134 134 134]]]
[[[ True  True  True]
  [ True  True  True]
  [ True  True  True]
  ..., 
  [ True  True  True]
  [ True  True  True]
  [ True  True  True]]

 [[ True  True  True]
  [ True  True  True]
  [ True  True  True]
  ..., 
  [ True  True  True]
  [ True  True  True]
  [ True  True  True]]

 [[ True  True  True]
  [ True  True  True]
  [ True  True  True]
  ..., 
  [ True  True  True]
  [ True  True  True]
  [ True  True  True]]

 ..., 
 [[ True  True  True]
  [ True  True  True]
  [ True  True  True]
  ..., 
  [ True  True  True]
  [ True  True  True]
  [ True  True  True]]

 [[ True  True  True]
  [ True  True  True]
  [ True  True  True]
  ..., 
  [ True  True  True]
  [ True  True  True]
  [ True  True  True]]

 [[ True  True  True]
  [ True  True  True]
  [ True  True  True]
  ..., 
  [ True  True  True]
  [ True  True  True]
  [ True  True  True]]]
(256, 256, 3)
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
0.704061302682

Answer 1

骰子系数等于图像元素数+图像2的交集元素数的2倍，在您的情况下，函数sum不给您元素数而是总和，只是由于 numpy 的逻辑交集不会为您提供相等的像素（请参阅上面的文档），我建议您像这样修改代码：

def dice_coef(img, img2):
        if img.shape != img2.shape:
            raise ValueError("Shape mismatch: img and img2 must have to be of the same shape.")
        else:
            
            lenIntersection=0
            
            for i in range(img.shape[0]):
                for j in range(img.shape[1]):
                    if ( np.array_equal(img[i][j],img2[i][j]) ):
                        lenIntersection+=1
             
            lenimg=img.shape[0]*img.shape[1]
            lenimg2=img2.shape[0]*img2.shape[1]  
            value = (2. * lenIntersection  / (lenimg + lenimg2))
        return value

计算相同大小的两个随机图像之间的骰子系数

1 个答案: