Spaces:
Build error
Build error
init
Browse files- app.py +155 -0
- modnet.py +94 -0
- requirements.txt +8 -0
app.py
ADDED
@@ -0,0 +1,155 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
|
3 |
+
import gradio as gr
|
4 |
+
|
5 |
+
import sys
|
6 |
+
sys.path.insert(0, 'U-2-Net')
|
7 |
+
|
8 |
+
from skimage import io, transform
|
9 |
+
import torch
|
10 |
+
import torchvision
|
11 |
+
from torch.autograd import Variable
|
12 |
+
import torch.nn as nn
|
13 |
+
import torch.nn.functional as F
|
14 |
+
from torch.utils.data import Dataset, DataLoader
|
15 |
+
from torchvision import transforms#, utils
|
16 |
+
# import torch.optim as optim
|
17 |
+
|
18 |
+
import numpy as np
|
19 |
+
from PIL import Image
|
20 |
+
import glob
|
21 |
+
|
22 |
+
from data_loader import RescaleT
|
23 |
+
from data_loader import ToTensor
|
24 |
+
from data_loader import ToTensorLab
|
25 |
+
from data_loader import SalObjDataset
|
26 |
+
|
27 |
+
from model import U2NET # full size version 173.6 MB
|
28 |
+
from model import U2NETP # small version u2net 4.7 MB
|
29 |
+
|
30 |
+
from modnet import ModNet
|
31 |
+
import huggingface_hub
|
32 |
+
|
33 |
+
# normalize the predicted SOD probability map
|
34 |
+
def normPRED(d):
|
35 |
+
ma = torch.max(d)
|
36 |
+
mi = torch.min(d)
|
37 |
+
|
38 |
+
dn = (d-mi)/(ma-mi)
|
39 |
+
|
40 |
+
return dn
|
41 |
+
def save_output(image_name,pred,d_dir):
|
42 |
+
predict = pred
|
43 |
+
predict = predict.squeeze()
|
44 |
+
predict_np = predict.cpu().data.numpy()
|
45 |
+
|
46 |
+
im = Image.fromarray(predict_np*255).convert('RGB')
|
47 |
+
img_name = image_name.split(os.sep)[-1]
|
48 |
+
image = io.imread(image_name)
|
49 |
+
imo = im.resize((image.shape[1],image.shape[0]),resample=Image.BILINEAR)
|
50 |
+
|
51 |
+
pb_np = np.array(imo)
|
52 |
+
|
53 |
+
aaa = img_name.split(".")
|
54 |
+
bbb = aaa[0:-1]
|
55 |
+
imidx = bbb[0]
|
56 |
+
for i in range(1,len(bbb)):
|
57 |
+
imidx = imidx + "." + bbb[i]
|
58 |
+
|
59 |
+
imo.save(d_dir+'/'+imidx+'.png')
|
60 |
+
return d_dir+'/'+imidx+'.png'
|
61 |
+
|
62 |
+
|
63 |
+
|
64 |
+
modnet_path = huggingface_hub.hf_hub_download('hylee/apdrawing_model',
|
65 |
+
'modnet.onnx',
|
66 |
+
force_filename='modnet.onnx')
|
67 |
+
modnet = ModNet(modnet_path)
|
68 |
+
|
69 |
+
# --------- 1. get image path and name ---------
|
70 |
+
model_name='u2net_portrait'#u2netp
|
71 |
+
|
72 |
+
|
73 |
+
image_dir = 'portrait_im'
|
74 |
+
prediction_dir = 'portrait_results'
|
75 |
+
if(not os.path.exists(prediction_dir)):
|
76 |
+
os.mkdir(prediction_dir)
|
77 |
+
|
78 |
+
model_dir = os.path.join(os.path.abspath(os.path.dirname(__file__)), 'U-2-Net/saved_models/u2net_portrait/u2net_portrait.pth')
|
79 |
+
|
80 |
+
|
81 |
+
# --------- 3. model define ---------
|
82 |
+
|
83 |
+
print("...load U2NET---173.6 MB")
|
84 |
+
net = U2NET(3,1)
|
85 |
+
|
86 |
+
net.load_state_dict(torch.load(model_dir, map_location='cpu'))
|
87 |
+
# if torch.cuda.is_available():
|
88 |
+
# net.cuda()
|
89 |
+
net.eval()
|
90 |
+
|
91 |
+
|
92 |
+
def process(im):
|
93 |
+
image = modnet.segment(im.name)
|
94 |
+
im_path = os.path.abspath(os.path.basename(im.name))
|
95 |
+
Image.fromarray(np.uint8(image)).save(im_path)
|
96 |
+
|
97 |
+
img_name_list = [im_path]
|
98 |
+
print("Number of images: ", len(img_name_list))
|
99 |
+
# --------- 2. dataloader ---------
|
100 |
+
# 1. dataloader
|
101 |
+
test_salobj_dataset = SalObjDataset(img_name_list=img_name_list,
|
102 |
+
lbl_name_list=[],
|
103 |
+
transform=transforms.Compose([RescaleT(512),
|
104 |
+
ToTensorLab(flag=0)])
|
105 |
+
)
|
106 |
+
test_salobj_dataloader = DataLoader(test_salobj_dataset,
|
107 |
+
batch_size=1,
|
108 |
+
shuffle=False,
|
109 |
+
num_workers=1)
|
110 |
+
|
111 |
+
results = []
|
112 |
+
# --------- 4. inference for each image ---------
|
113 |
+
for i_test, data_test in enumerate(test_salobj_dataloader):
|
114 |
+
|
115 |
+
print("inferencing:", img_name_list[i_test].split(os.sep)[-1])
|
116 |
+
|
117 |
+
inputs_test = data_test['image']
|
118 |
+
inputs_test = inputs_test.type(torch.FloatTensor)
|
119 |
+
|
120 |
+
# if torch.cuda.is_available():
|
121 |
+
# inputs_test = Variable(inputs_test.cuda())
|
122 |
+
# else:
|
123 |
+
inputs_test = Variable(inputs_test)
|
124 |
+
|
125 |
+
d1, d2, d3, d4, d5, d6, d7 = net(inputs_test)
|
126 |
+
|
127 |
+
# normalization
|
128 |
+
pred = 1.0 - d1[:, 0, :, :]
|
129 |
+
pred = normPRED(pred)
|
130 |
+
|
131 |
+
# save results to test_results folder
|
132 |
+
results.append(save_output(img_name_list[i_test], pred, prediction_dir))
|
133 |
+
|
134 |
+
del d1, d2, d3, d4, d5, d6, d7
|
135 |
+
|
136 |
+
print(results)
|
137 |
+
|
138 |
+
return Image.open(results[0])
|
139 |
+
|
140 |
+
title = "U-2-Net"
|
141 |
+
description = "Gradio demo for U-2-Net, https://github.com/xuebinqin/U-2-Net"
|
142 |
+
article = ""
|
143 |
+
|
144 |
+
gr.Interface(
|
145 |
+
process,
|
146 |
+
[gr.inputs.Image(type="file", label="Input")
|
147 |
+
],
|
148 |
+
[gr.outputs.Image(type="pil", label="Output")],
|
149 |
+
title=title,
|
150 |
+
description=description,
|
151 |
+
article=article,
|
152 |
+
examples=[],
|
153 |
+
allow_flagging=False,
|
154 |
+
allow_screenshot=False
|
155 |
+
).launch(enable_queue=True,cache_examples=True)
|
modnet.py
ADDED
@@ -0,0 +1,94 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import cv2
|
3 |
+
import argparse
|
4 |
+
import numpy as np
|
5 |
+
from PIL import Image
|
6 |
+
|
7 |
+
import onnx
|
8 |
+
import onnxruntime
|
9 |
+
|
10 |
+
|
11 |
+
class ModNet:
|
12 |
+
|
13 |
+
def __init__(self, model_path):
|
14 |
+
# Initialize session and get prediction
|
15 |
+
self.session = onnxruntime.InferenceSession(model_path, None)
|
16 |
+
|
17 |
+
# Get x_scale_factor & y_scale_factor to resize image
|
18 |
+
def get_scale_factor(self, im_h, im_w, ref_size):
|
19 |
+
|
20 |
+
if max(im_h, im_w) < ref_size or min(im_h, im_w) > ref_size:
|
21 |
+
if im_w >= im_h:
|
22 |
+
im_rh = ref_size
|
23 |
+
im_rw = int(im_w / im_h * ref_size)
|
24 |
+
elif im_w < im_h:
|
25 |
+
im_rw = ref_size
|
26 |
+
im_rh = int(im_h / im_w * ref_size)
|
27 |
+
else:
|
28 |
+
im_rh = im_h
|
29 |
+
im_rw = im_w
|
30 |
+
|
31 |
+
im_rw = im_rw - im_rw % 32
|
32 |
+
im_rh = im_rh - im_rh % 32
|
33 |
+
|
34 |
+
x_scale_factor = im_rw / im_w
|
35 |
+
y_scale_factor = im_rh / im_h
|
36 |
+
|
37 |
+
return x_scale_factor, y_scale_factor
|
38 |
+
|
39 |
+
def segment(self, image_path):
|
40 |
+
ref_size = 512
|
41 |
+
##############################################
|
42 |
+
# Main Inference part
|
43 |
+
##############################################
|
44 |
+
|
45 |
+
# read image
|
46 |
+
im = cv2.imread(image_path)
|
47 |
+
im = cv2.cvtColor(im, cv2.COLOR_BGR2RGB)
|
48 |
+
|
49 |
+
# unify image channels to 3
|
50 |
+
if len(im.shape) == 2:
|
51 |
+
im = im[:, :, None]
|
52 |
+
if im.shape[2] == 1:
|
53 |
+
im = np.repeat(im, 3, axis=2)
|
54 |
+
elif im.shape[2] == 4:
|
55 |
+
im = im[:, :, 0:3]
|
56 |
+
|
57 |
+
# normalize values to scale it between -1 to 1
|
58 |
+
im = (im - 127.5) / 127.5
|
59 |
+
|
60 |
+
im_h, im_w, im_c = im.shape
|
61 |
+
x, y = self.get_scale_factor(im_h, im_w, ref_size)
|
62 |
+
|
63 |
+
# resize image
|
64 |
+
im = cv2.resize(im, None, fx=x, fy=y, interpolation=cv2.INTER_AREA)
|
65 |
+
|
66 |
+
# prepare input shape
|
67 |
+
im = np.transpose(im)
|
68 |
+
im = np.swapaxes(im, 1, 2)
|
69 |
+
im = np.expand_dims(im, axis=0).astype('float32')
|
70 |
+
|
71 |
+
input_name = self.session.get_inputs()[0].name
|
72 |
+
output_name = self.session.get_outputs()[0].name
|
73 |
+
result = self.session.run([output_name], {input_name: im})
|
74 |
+
|
75 |
+
# refine matte
|
76 |
+
matte = (np.squeeze(result[0]) * 255).astype('uint8')
|
77 |
+
matte = cv2.resize(matte, dsize=(im_w, im_h), interpolation=cv2.INTER_AREA)
|
78 |
+
|
79 |
+
# obtain predicted foreground
|
80 |
+
image = cv2.imread(image_path)
|
81 |
+
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
|
82 |
+
|
83 |
+
if len(image.shape) == 2:
|
84 |
+
image = image[:, :, None]
|
85 |
+
if image.shape[2] == 1:
|
86 |
+
image = np.repeat(image, 3, axis=2)
|
87 |
+
elif image.shape[2] == 4:
|
88 |
+
image = image[:, :, 0:3]
|
89 |
+
matte = np.repeat(np.asarray(matte)[:, :, None], 3, axis=2) / 255
|
90 |
+
foreground = image * matte + np.full(image.shape, 255) * (1 - matte)
|
91 |
+
|
92 |
+
return foreground
|
93 |
+
|
94 |
+
|
requirements.txt
ADDED
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
numpy
|
2 |
+
scikit-image
|
3 |
+
torch
|
4 |
+
torchvision
|
5 |
+
pillow
|
6 |
+
opencv-python-headless
|
7 |
+
onnx==1.8.1
|
8 |
+
onnxruntime==1.6.0
|