File size: 1,945 Bytes
447ff7c |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 |
import json
import cv2
import numpy as np
import os
from torch.utils.data import Dataset
from PIL import Image
import cv2
from .data_utils import *
from .base import BaseDataset
import albumentations as A
class DresscodeDataset(BaseDataset):
def __init__(self, image_dir):
self.image_root = image_dir
self.data = os.listdir(self.image_root)
self.size = (512,512)
self.clip_size = (224,224)
self.dynamic = 2
def __len__(self):
return 20000
def check_region_size(self, image, yyxx, ratio, mode = 'max'):
pass_flag = True
H,W = image.shape[0], image.shape[1]
H,W = H * ratio, W * ratio
y1,y2,x1,x2 = yyxx
h,w = y2-y1,x2-x1
if mode == 'max':
if h > H and w > W:
pass_flag = False
elif mode == 'min':
if h < H and w < W:
pass_flag = False
return pass_flag
def get_sample(self, idx):
tar_mask_path = os.path.join(self.image_root, self.data[idx])
tar_image_path = tar_mask_path.replace('label_maps/','images/').replace('_4.png','_0.jpg')
ref_image_path = tar_mask_path.replace('label_maps/','images/').replace('_4.png','_1.jpg')
# Read Image and Mask
ref_image = cv2.imread(ref_image_path)
ref_image = cv2.cvtColor(ref_image, cv2.COLOR_BGR2RGB)
tar_image = cv2.imread(tar_image_path)
tar_image = cv2.cvtColor(tar_image, cv2.COLOR_BGR2RGB)
ref_mask = (ref_image < 240).astype(np.uint8)[:,:,0]
tar_mask = Image.open(tar_mask_path ).convert('P')
tar_mask= np.array(tar_mask)
tar_mask = tar_mask == 4
item_with_collage = self.process_pairs(ref_image, ref_mask, tar_image, tar_mask, max_ratio = 1.0)
sampled_time_steps = self.sample_timestep()
item_with_collage['time_steps'] = sampled_time_steps
return item_with_collage
|