深度学习之图像分割结果可视化

图像分割结果一般是二值图像(黑白)，为了更好的查看分割结果我们将正确分割显示为白色，背景为黑色，错误检测为红色，漏检的为绿色。

对比结果如下：

基于python的实现代码如下：

import os
import cv2

def clmap(v, k, upBound): #mul and clamp
    val = v * k
    if val > upBound:
        return upBound
    else:
        return val
source_path1 =r'G:\2\...' # <----------标签源文件路径
source_path2 =r'G:\2\...' # <----------结果源文件路径
target_path = r'G:\2\...' # <----------输出目标文件路径


image_list = os.listdir(source_path1)  # 获得文件名

i = 0
for file in image_list:
    print(source_path1 + "\\" + file)
    img_2 = cv2.imread(source_path1 + "\\" + file)  # 循环读取图片
    img_1 = cv2.imread(source_path2 +  "\\" + file)  # 循环读取图片
    img_3 = cv2.imread(r'G:\2\...')  # 纯黑色背景图片
   
    dif = img_3.copy()
    show_dif = dif.copy()  # dif image for show only
    
    width = img_3.shape[0]  # get width
    height = img_3.shape[1]  # get height
    
    for i in range(width):
        for j in range(height):
            diff = int(img_1[i, j][0]) - int(img_2[i, j][0])
            if diff < 0:
                show_dif[i, j] = [255, clmap(abs(diff), 10, 255), 255]
                
            elif diff > 0:
                show_dif[i, j] = [255, 255, clmap(abs(diff), 10, 255)]

创作不易，还请支持: