How to calculate IoU? about medical-transformer

[over-star]

generate mask image is 8-bit, please ensure your image in labelcol also is 8-bit,then calculate IoU like :

import os
import numpy as np
from sklearn.metrics import jaccard_score, f1_score
from PIL import Image,ImageDraw

N = 328
st = 0
Fsc = []
MIU = []
PA = []
bestfsc = 0
bestmiu = 0
bestpa = 0
bestep = 0
folder = './datasets/CrackForest/value/labelcol'
listinfo = os.listdir(folder)

f1, iou = [], []
for i in listinfo:
    I = np.array(Image.open('./datasets/CrackForest/value/labelcol/' + i))
    I = np.where(I >= 127, 1, 0)
    mask = i.replace("png", "jpg")
    L = np.array(Image.open('./datasets/CrackForest/value/result/' + mask))
    L = np.where(L >= 127, 1, 0)
    y_true, y_pred = I.flatten(), L.flatten()
    f1.append(f1_score(y_true, y_pred))
    iou.append(jaccard_score(y_true, y_pred))
print(f1)

[jeya-maria-jose]

You can use the performancemetrics_monuseg.m code provided in the repo !