-
Notifications
You must be signed in to change notification settings - Fork 0
/
inference.py
63 lines (58 loc) · 2.59 KB
/
inference.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
63
import torch
import cv2
import torch.nn.functional as F
import matplotlib.pyplot as plt
import numpy as np
from main import SuperRes
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
print(device)
test_dir = "test_images"
def upsample(img):
with torch.no_grad():
img = cv2.cvtColor(img, cv2.COLOR_BGR2YCR_CB)
img = torch.tensor(img).to(device).unsqueeze(0).permute(0, 3, 1, 2)
y = img[:,[0],:,:]
# img = img.permute(1, 0, 2, 3)
upsampled_cc = F.interpolate(img[:,1:,:,:], scale_factor=2, mode='bicubic')
upsampled_y = model(y)
upsampled = torch.cat([upsampled_y, upsampled_cc], dim=1).squeeze().permute(1,2,0).cpu().numpy()
# upsampled = F.interpolate(img, scale_factor=2, mode='bicubic').squeeze().permute(1, 2, 0).cpu().numpy()
# upsampled = model(img).squeeze().permute(1, 2, 0).cpu().numpy()
upsampled = cv2.cvtColor(upsampled, cv2.COLOR_YCR_CB2BGR)
return upsampled
def upsample_naive(img):
with torch.no_grad():
img = cv2.cvtColor(img, cv2.COLOR_BGR2YCR_CB)
img = torch.tensor(img).to(device).unsqueeze(0).permute(0, 3, 1, 2)
upsampled = F.interpolate(img, scale_factor=2, mode='bicubic').squeeze().permute(1,2,0).cpu().numpy()
upsampled = cv2.cvtColor(upsampled, cv2.COLOR_YCR_CB2BGR)
return upsampled
if __name__ == "__main__":
PT_PATH = "saved.pt"
model = torch.load(PT_PATH)
custom_img = [
cv2.imread("test_images/green-maple-leaf.jpg"),
cv2.imread("test_images/home-office.jpg"),
cv2.imread("test_images/martin-luther-king.jpg"),
cv2.imread("test_images/mount-rushmore.jpg"),
cv2.imread("test_images/salisbury-cathedral.jpg")]
downsampled =[cv2.resize(img.astype(np.float32)/255.0, (40,40)) for img in custom_img ]
upsampled = [upsample(img) for img in downsampled]
upsampled_naive = [upsample_naive(img) for img in downsampled]
n = 5
plt.figure(figsize=(20, 10))
for i in range(n):
ax = plt.subplot(3, n, i+1)
plt.imshow(downsampled[i][:,:,::-1])
ax.get_xaxis().set_visible(False)
ax.get_yaxis().set_visible(False)
ax = plt.subplot(3, n, i+1+n)
plt.imshow(upsampled[i][:,:,::-1])
ax.get_xaxis().set_visible(False)
ax.get_yaxis().set_visible(False)
ax = plt.subplot(3, n, i+1+2*n)
plt.imshow(upsampled_naive[i][:,:,::-1])
ax.get_xaxis().set_visible(False)
ax.get_yaxis().set_visible(False)
# plt.show()
plt.savefig('inference.png')