-
Notifications
You must be signed in to change notification settings - Fork 0
/
test_image_directory.py
63 lines (46 loc) · 1.88 KB
/
test_image_directory.py
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 argparse
import os
import numpy as np
import torch.backends.cudnn as cudnn
import torch.utils.data.distributed
import torchvision.transforms as transforms
from PIL import Image
from vdsr import VDSR
parser = argparse.ArgumentParser(description="Gradient Variance loss for structure-enhanced super-resolution")
parser.add_argument("--dataroot", type=str, required=True, help="The directory path where the image needs ")
parser.add_argument("--scale-factor", type=int, default=2, choices=[2, 3, 4, 8],
help="Image scaling ratio. (default: 2).")
parser.add_argument("--weights", type=str, required=True, help="path to the model weights")
parser.add_argument("--cuda", action="store_true", help="Enables cuda")
args = parser.parse_args()
print(args)
try:
os.makedirs("result")
except OSError:
pass
cudnn.benchmark = True
if torch.cuda.is_available() and not args.cuda:
print("WARNING: CUDA device available, consider run with --cuda")
device = torch.device("cuda:0" if args.cuda else "cpu")
# define model
model = VDSR().to(device)
# Load pretrained weights
model.load_state_dict(torch.load(args.weights, map_location=device))
dataroot = args.dataroot
scale_factor = args.scale_factor
for filename in os.listdir(dataroot):
# open image and upscale
image = Image.open(f"{dataroot}/{filename}")
image_width = int(image.size[0] * scale_factor)
image_height = int(image.size[1] * scale_factor)
image = image.resize((image_width, image_height), Image.BICUBIC)
preprocess = transforms.ToTensor()
inputs = preprocess(image).view(1, -1, image.size[1], image.size[0])
inputs = inputs.to(device)
out = model(inputs)
out = out.cpu()
out_image = out[0].detach().numpy()
out_image *= 255.0
out_image = out_image.clip(0, 255).transpose(1, 2, 0)
out_image = Image.fromarray(np.uint8(out_image))
out_image.save(f"result/{filename}")