前两天进行一个分割项目，模型训练完成后需要对分割效果可视化，特此记录两种方法。

---

  示例中用到的原图、label和mask如下所示：

1. 标签label填充，预测mask勾线

  第一种是标签label填充样覆盖，预测mask采用勾线描边的方式，勾线描边利用 $cv2$ 中的 $cv2.threshold$、$cv2.findContours$ 和 $cv2.drawContours$ 函数。

import cv2
import matplotlib.pyplot as plt
import numpy as np
from PIL import Image
from torchvision.transforms import transforms

def show_mask(mask, ax, random_color=False):
    if random_color:
        color = np.concatenate([np.random.random(3), np.array([0.6])], axis=0)
    else:
        color = np.array([30 / 255, 144 / 255, 255 / 255, 0.5])
    h, w = mask.shape[-2:]
    mask_image = mask.reshape(h, w, 1) * color.reshape(1, 1, -1)
    ax.imshow(mask_image)
    return mask_image

image = cv2.imread('./data/0.png') 
label = Image.open("./data/0_gt.jpg")
mask = Image.open("./data/0_pred.jpg")

h, w, _ = image.shape

label = transforms.Resize((h,w))(label)
label = np.uint8(label)
label = 255 * (label > 127)
label = label == 255

mask = transforms.Resize((h,w))(mask)
mask = np.uint8(mask)

ret_mask, thresh_mask = cv2.threshold(mask, 127, 255, 0)
contours_mask, hierarchy_mask = cv2.findContours(thresh_mask, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)

for cnt_mask in contours_mask:
    cv2.drawContours(image, [cnt_mask], 0, (0,255,0), 2)

plt.figure(dpi=300)
plt.imshow(image)
show_mask(label, plt.gca())
plt.axis('off')
plt.savefig("output_image.jpg")

效果图：

2. 标签label勾线，预测mask勾线

  第二种是利用不用的颜色均采用勾线描边的方式， $cv2.drawContours$ 自带 hold on 功能，只需两个循环就搞定啦~

import cv2
import numpy as np
from PIL import Image, ImageDraw
from torchvision.transforms import transforms

image = cv2.imread('./data/0.png') 
label = Image.open("./data/0_gt.jpg")
mask = Image.open("./data/0_pred.jpg")

h, w, _ = image.shape

label = transforms.Resize((h,w))(label)
label = np.uint8(label)

mask = transforms.Resize((h,w))(mask)
mask = np.uint8(mask)

ret_label, thresh_label = cv2.threshold(label, 127, 255, 0)
contours_label, hierarchy_label = cv2.findContours(thresh_label, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)

ret_mask, thresh_mask = cv2.threshold(mask, 127, 255, 0)
contours_mask, hierarchy_mask = cv2.findContours(thresh_mask, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)

for cnt_label in contours_label:
    cv2.drawContours(image, [cnt_label], 0, (0,0,255), 2)
    
for cnt_mask in contours_mask:
    cv2.drawContours(image, [cnt_mask], 0, (0,255,0), 2)

cv2.imwrite('output_image.jpg', image)

效果图：