Update app.py
Browse files
app.py
CHANGED
@@ -1,131 +1,85 @@
|
|
1 |
import os
|
2 |
-
os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID"
|
3 |
-
|
4 |
import gradio as gr
|
5 |
import torch
|
6 |
import cv2
|
7 |
import numpy as np
|
8 |
-
from preprocess import unsharp_masking
|
9 |
-
import glob
|
10 |
import time
|
11 |
|
12 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
13 |
|
14 |
-
|
15 |
-
|
16 |
-
)
|
17 |
-
|
18 |
-
def ordenar_arquivos(img, modelo):
|
19 |
-
ori = img.copy()
|
20 |
img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
|
21 |
-
h, w = img.shape
|
22 |
-
img_out = preprocessamento(img, modelo)
|
23 |
-
return img_out, h, w, img, ori
|
24 |
-
|
25 |
-
def preprocessamento(img, modelo='SE-RegUNet 4GF'):
|
26 |
img = cv2.resize(img, (512, 512))
|
27 |
-
img =
|
28 |
-
|
29 |
-
|
30 |
-
img_out = np.expand_dims(img, axis=0)
|
31 |
-
elif modelo == 'SE-RegUNet 4GF':
|
32 |
-
clahe1 = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8, 8))
|
33 |
-
clahe2 = cv2.createCLAHE(clipLimit=8.0, tileGridSize=(8, 8))
|
34 |
-
image1 = clahe1.apply(img)
|
35 |
-
image2 = clahe2.apply(img)
|
36 |
-
img = np.float32((img - img.min()) / (img.max() - img.min() + 1e-6))
|
37 |
-
image1 = np.float32((image1 - image1.min()) / (image1.max() - image1.min() + 1e-6))
|
38 |
-
image2 = np.float32((image2 - image2.min()) / (image2.max() - image2.min() + 1e-6))
|
39 |
-
img_out = np.stack((img, image1, image2), axis=0)
|
40 |
-
else:
|
41 |
-
clahe1 = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8, 8))
|
42 |
-
image1 = clahe1.apply(img)
|
43 |
-
image1 = np.float32((image1 - image1.min()) / (image1.max() - image1.min() + 1e-6))
|
44 |
-
img_out = np.stack((image1,) * 3, axis=0)
|
45 |
-
return img_out
|
46 |
-
|
47 |
-
def processar_imagem_de_entrada(img, modelo, pipe):
|
48 |
-
img = img.copy()
|
49 |
-
pipe = pipe.to(device).eval()
|
50 |
-
start = time.time()
|
51 |
-
img, h, w, ori_gray, ori = ordenar_arquivos(img, modelo)
|
52 |
-
img = torch.FloatTensor(img).unsqueeze(0).to(device)
|
53 |
-
with torch.no_grad():
|
54 |
-
if modelo == 'AngioNet':
|
55 |
-
img = torch.cat([img, img], dim=0)
|
56 |
-
logit = np.round(torch.softmax(pipe.forward(img), dim=1).detach().cpu().numpy()[0, 0]).astype(np.uint8)
|
57 |
-
spent = time.time() - start
|
58 |
-
spent = f"{spent:.3f} segundos"
|
59 |
-
|
60 |
-
if h != 512 or w != 512:
|
61 |
-
logit = cv2.resize(logit, (h, w))
|
62 |
-
|
63 |
-
logit = logit.astype(bool)
|
64 |
-
img_out = ori.copy()
|
65 |
-
img_out[logit, 0] = 255
|
66 |
-
return spent, img_out
|
67 |
|
68 |
-
|
69 |
-
|
70 |
-
|
71 |
-
'SE-RegUNet 16GF': torch.jit.load('./model/SERegUNet16GF.pt'),
|
72 |
-
'AngioNet': torch.jit.load('./model/AngioNet.pt'),
|
73 |
-
'EffUNet++ B5': torch.jit.load('./model/EffUNetppb5.pt'),
|
74 |
-
'Reg-SA-UNet++': torch.jit.load('./model/RegSAUnetpp.pt'),
|
75 |
-
'UNet3+': torch.jit.load('./model/UNet3plus.pt'),
|
76 |
-
}
|
77 |
|
78 |
-
def
|
79 |
-
|
80 |
-
|
81 |
-
|
82 |
-
|
83 |
-
|
84 |
-
"""
|
85 |
-
Checks if the angiogram has disease based on the segmentation.
|
86 |
-
Args:
|
87 |
-
img_out: The segmented angiogram.
|
88 |
-
Returns:
|
89 |
-
True if the angiogram has disease, False otherwise.
|
90 |
-
"""
|
91 |
-
percentage_of_vessels = np.sum(img_out) / (img_out.shape[0] * img_out.shape[1])
|
92 |
-
if percentage_of_vessels > 0.5:
|
93 |
-
return True
|
94 |
-
else:
|
95 |
-
return False
|
96 |
|
97 |
-
|
|
|
|
|
98 |
|
99 |
-
|
100 |
-
if
|
101 |
-
|
102 |
-
|
103 |
)
|
104 |
else:
|
105 |
-
|
106 |
-
|
107 |
-
|
108 |
-
|
109 |
-
|
110 |
-
|
111 |
-
|
112 |
-
|
113 |
-
|
114 |
-
|
115 |
-
|
116 |
-
|
117 |
-
|
118 |
-
|
119 |
-
|
120 |
-
|
121 |
-
|
122 |
-
|
123 |
-
|
124 |
-
|
125 |
-
|
126 |
-
|
127 |
-
|
128 |
-
|
129 |
-
|
130 |
-
|
131 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
import os
|
|
|
|
|
2 |
import gradio as gr
|
3 |
import torch
|
4 |
import cv2
|
5 |
import numpy as np
|
|
|
|
|
6 |
import time
|
7 |
|
8 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
9 |
|
10 |
+
def preprocess_image(img, model_name):
|
11 |
+
# Preprocess the input image based on the selected model
|
|
|
|
|
|
|
|
|
12 |
img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
|
|
|
|
|
|
|
|
|
|
|
13 |
img = cv2.resize(img, (512, 512))
|
14 |
+
img = cv2.GaussianBlur(img, (0, 0), 1.0)
|
15 |
+
img = img.astype(np.float32) / 255.0
|
16 |
+
return img
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
17 |
|
18 |
+
def load_model(model_path):
|
19 |
+
model = torch.jit.load(model_path)
|
20 |
+
return model
|
|
|
|
|
|
|
|
|
|
|
|
|
21 |
|
22 |
+
def process_image(img, model):
|
23 |
+
img = np.expand_dims(img, axis=0)
|
24 |
+
img_tensor = torch.FloatTensor(img).unsqueeze(0).to(device)
|
25 |
+
with torch.no_grad():
|
26 |
+
logit = torch.softmax(model.forward(img_tensor), dim=1).detach().cpu().numpy()[0, 0]
|
27 |
+
return (logit > 0.5).astype(np.uint8)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
28 |
|
29 |
+
def has_disease(segmented_img):
|
30 |
+
percentage_of_vessels = np.sum(segmented_img) / (segmented_img.shape[0] * segmented_img.shape[1])
|
31 |
+
return percentage_of_vessels > 0.5
|
32 |
|
33 |
+
def explanation(has_disease_flag):
|
34 |
+
if has_disease_flag:
|
35 |
+
explanation_text = (
|
36 |
+
"O angiograma tem doença. Isso é determinado pela função `has_disease`, que calcula o percentual de vasos no angiograma segmentado. Se o percentual de vasos for maior que 50%, a função retorna \"true\", indicando que o angiograma tem doença. Caso contrário, a função retorna \"false\", indicando que o angiograma não tem doença."
|
37 |
)
|
38 |
else:
|
39 |
+
explanation_text = "O angiograma não tem doença."
|
40 |
+
return explanation_text
|
41 |
+
|
42 |
+
def process_input_image(input_img, model_name, model):
|
43 |
+
start_time = time.time()
|
44 |
+
|
45 |
+
preprocessed_img = preprocess_image(input_img, model_name)
|
46 |
+
processed_img = process_image(preprocessed_img, model)
|
47 |
+
disease_flag = has_disease(processed_img)
|
48 |
+
explanation_text = explanation(disease_flag)
|
49 |
+
|
50 |
+
elapsed_time = f"{time.time() - start_time:.3f} segundos"
|
51 |
+
return elapsed_time, processed_img, disease_flag, explanation_text
|
52 |
+
|
53 |
+
def main():
|
54 |
+
# Load models
|
55 |
+
model_paths = {
|
56 |
+
'SE-RegUNet 4GF': './model/SERegUNet4GF.pt',
|
57 |
+
# ... (other model paths here)
|
58 |
+
}
|
59 |
+
models = {model_name: load_model(model_path) for model_name, model_path in model_paths.items()}
|
60 |
+
|
61 |
+
# Create Gradio interface
|
62 |
+
gr_interface = gr.Interface(
|
63 |
+
fn=process_input_image,
|
64 |
+
inputs=[
|
65 |
+
gr.inputs.Image(label="Angiograma:", shape=(512, 512)),
|
66 |
+
gr.inputs.Dropdown(list(model_paths.keys()), label='Modelo', default='SE-RegUNet 4GF'),
|
67 |
+
],
|
68 |
+
outputs=[
|
69 |
+
gr.outputs.Label(label="Tempo decorrido"),
|
70 |
+
gr.outputs.Image(type="numpy", label="Imagem de Saída"),
|
71 |
+
gr.outputs.Label(label="Possui Doença?"),
|
72 |
+
gr.outputs.Label(label="Explicação"),
|
73 |
+
],
|
74 |
+
title="Segmentação de Angiograma Coronariano",
|
75 |
+
description="Esta aplicação segmenta angiogramas coronarianos usando modelos de segmentação pré-treinados. Faça o upload de uma imagem de angiograma e selecione um modelo para visualizar o resultado da segmentação.\n\nSelecione uma imagem de angiograma coronariano e um modelo de segmentação no painel à esquerda.",
|
76 |
+
theme="default",
|
77 |
+
layout="vertical",
|
78 |
+
allow_flagging=False,
|
79 |
+
)
|
80 |
+
|
81 |
+
# Launch Gradio interface
|
82 |
+
gr_interface.launch()
|
83 |
+
|
84 |
+
if __name__ == "__main__":
|
85 |
+
main()
|