dice_test

import glob
import os
import cv2
import numpy as np

def dice(pre_img,mask_img):
    # smooth = 1e-5
    s = []

    s1 = cv2.imread(pre_img,0)
    pre_img = cv2.resize(s1, (256, 256))
    pre_img[pre_img <125]=0
    pre_img[pre_img > 125] = 1
    s1 = pre_img

    mask_img = cv2.imread(mask_img,0)
    mask_img[mask_img < 125] = 0
    mask_img[mask_img >= 125] = 1
    s2 = mask_img

    # print(pre_img.shape)
    row,col = s2.shape[0],s2.shape[1]
    for r in range(row):
        for c in range(col):
            if s1[r][c] == s2[r][c]:
                s.append(s2[r][c])
    # print(s)
    intersection = np.linalg.norm(s)
    output = np.linalg.norm(s1.flatten()) + np.linalg.norm(s2.flatten())
    return (2 * intersection) / (output)


pre_path = '/home/zhengsc/Desktop/pytorch-nested-unet/outputs/colon_train_UNet_woDS_double_up/0/'
mask_path = '/home/zhengsc/Desktop/pytorch-nested-unet/inputs/colon_test/masks/0/'
pre_imgs = glob.glob(pre_path+'*.jpg')
DICE = []
# mask_imgs = glob.glob(mask_path+'*.jpg')
for pre_img in pre_imgs:
    name = pre_img[84:]
    mask_img = os.path.join(mask_path,name)
    Dice = dice(pre_img,mask_img)
    DICE.append(Dice)

dice_mean = np.mean(DICE)
dice_var = np.var(DICE)
std = np.std(DICE,ddof=1)
print(dice_mean,dice_var,std)