Update app.py
Browse files
app.py
CHANGED
@@ -8,18 +8,59 @@ os.system("pip3 install torch")
|
|
8 |
os.system("pip3 install collections")
|
9 |
os.system("pip3 install torchvision")
|
10 |
os.system("pip3 install einops")
|
|
|
11 |
#os.system("pip3 install argparse")
|
12 |
-
|
13 |
from PIL import Image
|
14 |
import torch
|
15 |
from torchvision import transforms
|
16 |
from model_video import build_model
|
17 |
import numpy as np
|
18 |
import collections
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
19 |
#import argparse
|
20 |
device='cuda:0'
|
21 |
net = build_model(device).to(device)
|
22 |
-
|
23 |
model_path = 'image_best.pth'
|
24 |
print(model_path)
|
25 |
weight=torch.load(model_path,map_location=torch.device(device))
|
@@ -44,8 +85,10 @@ def test(gpu_id, net, img_list, group_size, img_size):
|
|
44 |
for i in range(5):
|
45 |
group_img[i]=img_transform(Image.fromarray(img_list[i]))
|
46 |
_,pred_mask=net(group_img)
|
47 |
-
|
48 |
-
|
|
|
|
|
49 |
#w, h = 224,224#Image.open(image_list[i][j]).size
|
50 |
#result = result.resize((w, h), Image.BILINEAR)
|
51 |
#result.convert('L').save('0.png')
|
|
|
8 |
os.system("pip3 install collections")
|
9 |
os.system("pip3 install torchvision")
|
10 |
os.system("pip3 install einops")
|
11 |
+
os.system("pip3 install pydensecrf")
|
12 |
#os.system("pip3 install argparse")
|
13 |
+
import pydensecrf.densecrf as dcrf
|
14 |
from PIL import Image
|
15 |
import torch
|
16 |
from torchvision import transforms
|
17 |
from model_video import build_model
|
18 |
import numpy as np
|
19 |
import collections
|
20 |
+
|
21 |
+
def crf_refine(img, annos):
|
22 |
+
def _sigmoid(x):
|
23 |
+
return 1 / (1 + np.exp(-x))
|
24 |
+
|
25 |
+
assert img.dtype == np.uint8
|
26 |
+
assert annos.dtype == np.uint8
|
27 |
+
assert img.shape[:2] == annos.shape
|
28 |
+
|
29 |
+
# img and annos should be np array with data type uint8
|
30 |
+
|
31 |
+
EPSILON = 1e-8
|
32 |
+
|
33 |
+
M = 2 # salient or not
|
34 |
+
tau = 1.05
|
35 |
+
# Setup the CRF model
|
36 |
+
d = dcrf.DenseCRF2D(img.shape[1], img.shape[0], M)
|
37 |
+
|
38 |
+
anno_norm = annos / 255.
|
39 |
+
|
40 |
+
n_energy = -np.log((1.0 - anno_norm + EPSILON)) / (tau * _sigmoid(1 - anno_norm))
|
41 |
+
p_energy = -np.log(anno_norm + EPSILON) / (tau * _sigmoid(anno_norm))
|
42 |
+
|
43 |
+
U = np.zeros((M, img.shape[0] * img.shape[1]), dtype='float32')
|
44 |
+
U[0, :] = n_energy.flatten()
|
45 |
+
U[1, :] = p_energy.flatten()
|
46 |
+
|
47 |
+
d.setUnaryEnergy(U)
|
48 |
+
|
49 |
+
d.addPairwiseGaussian(sxy=3, compat=3)
|
50 |
+
d.addPairwiseBilateral(sxy=60, srgb=5, rgbim=img, compat=5)
|
51 |
+
|
52 |
+
# Do the inference
|
53 |
+
infer = np.array(d.inference(1)).astype('float32')
|
54 |
+
res = infer[1, :]
|
55 |
+
|
56 |
+
res = res * 255
|
57 |
+
res = res.reshape(img.shape[:2])
|
58 |
+
return res.astype('uint8')
|
59 |
+
|
60 |
#import argparse
|
61 |
device='cuda:0'
|
62 |
net = build_model(device).to(device)
|
63 |
+
#net=torch.nn.DataParallel(net)
|
64 |
model_path = 'image_best.pth'
|
65 |
print(model_path)
|
66 |
weight=torch.load(model_path,map_location=torch.device(device))
|
|
|
85 |
for i in range(5):
|
86 |
group_img[i]=img_transform(Image.fromarray(img_list[i]))
|
87 |
_,pred_mask=net(group_img)
|
88 |
+
pred_mask=(pred_mask.detach().squeeze()*255).numpy().astype(np.uint8)
|
89 |
+
pred_mask=[crf_refine(img_list[i],pred_mask[i]) for i in range(5)]
|
90 |
+
print(pred_mask[0].shape)
|
91 |
+
result = [Image.fromarray((torch.from_numpy(pred_mask[i]).unsqueeze(2).repeat(1,1,3)).numpy()) for i in range(5)]
|
92 |
#w, h = 224,224#Image.open(image_list[i][j]).size
|
93 |
#result = result.resize((w, h), Image.BILINEAR)
|
94 |
#result.convert('L').save('0.png')
|