CelebAHQ dataset incompatibility

[Jaep0805]

The dataloader you've provided seem to be incompatible with the CelebAHQ dataset.
CelebAHQ dataset does not provide the segmentation mask as a whole, but rather in categories.
It is also divided into subfolders depending on the numbers of the image, from 0~14

[Jaep0805]

solved

[hanpiness]

I have encountered the same problem. May I ask how you resolved it? Thank you！

[Sanoojan]

import os
from PIL import Image
import numpy as np

celelbAHQ_label_list = ['skin', 'nose', 'eye_g', 'l_eye', 'r_eye','l_brow', 'r_brow', 'l_ear', 'r_ear', 'mouth','u_lip', 'l_lip', 'hair', 'hat', `'ear_r','neck_l',` 'neck', 'cloth']

faceParser_label_list_detailed = ['background', 'lip', 'eyebrows', 'eyes', 'hair', 'nose', 'skin', 'ears', 'belowface', 'mouth', 'eye_glass', 'ear_rings']

Dataset_maskPath='e4s/data/CelebAMask-HQ/CelebA-HQ-mask'
save_path='e4s/data/CelebAMask-HQ/CelebA-HQ-mask/Overall_mask'
if not os.path.exists(save_path):
    os.mkdir(save_path)
for i in range(30000):
    mask=np.zeros((512,512))
    for ind,cate in enumerate(celelbAHQ_label_list):
        
        if os.path.exists(os.path.join(Dataset_maskPath,"%d"%int(i/2000) ,'{0:0=5d}'.format(i)+'_'+cate+'.png')):
            im= Image.open(os.path.join(Dataset_maskPath,"%d"%int(i/2000) ,'{0:0=5d}'.format(i)+'_'+cate+'.png')).convert('L')
            im = np.equal(im, 255)
            mask[im]=ind+1

    cv2.imwrite(os.path.join(save_path,'%d'%i+'.png'),mask)

@hanpiness I created the mask like this to solve the problem. Hope it may help you.

[Jeriousman]

solved

How did you solve it?