-
Notifications
You must be signed in to change notification settings - Fork 1
/
inference_api.py
62 lines (49 loc) · 1.51 KB
/
inference_api.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
import torch
from network import FINED
import os
import numpy as np
from datasets import BSDS_Dataset
from torch.utils.data import DataLoader
from PIL import Image
import scipy.io as io
import torchvision
from os.path import join
import cv2
from PIL import Image
# CONFIG
os.environ["CUDA_VISIBLE_DEVICES"]="0"
weight_file = 'weights/final-model.pth'
def read_imagefile(file) -> Image.Image:
image = Image.open(BytesIO(file))
return image
def get_FINED_edge(im):
## Multiscale
scales = [0.5,1.0,1.5]
images = []
for scl in scales:
img_scale = cv2.resize(im, None, fx=scl, fy=scl, interpolation=cv2.INTER_LINEAR)
images.append(img_scale.transpose(2, 0, 1)) # (H x W x C) to (C x H x W)
# CREATE MODEL
model = FINED(pretrain=weight_file,isTrain=False)
model.cuda()
model.eval()
## FEED FORWARD
h, w, _ = im.shape
ms_fuse = np.zeros((h, w))
retn = None
with torch.no_grad():
for img in images:
img = img[np.newaxis, :, :, :]
img = torch.from_numpy(img)
img = img.cuda()
img = img.float()
out = model(img)
fuse = out[-1].squeeze().detach().cpu().numpy()
fuse = cv2.resize(fuse, (w, h), interpolation=cv2.INTER_LINEAR)
ms_fuse += fuse
ms_fuse /= len(scales)
retn = 255 - (ms_fuse * 255).astype(np.uint8)
filename = 'prediction'
result = Image.fromarray(retn)
result.save( "img/result_%s.png" % filename)
return retn