Upload app.py
Browse files
app.py
ADDED
@@ -0,0 +1,87 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
import numpy as np
|
3 |
+
import cv2
|
4 |
+
from mmseg.apis import init_model, inference_model
|
5 |
+
import torch
|
6 |
+
|
7 |
+
|
8 |
+
def process_single_img(img_bgr, model_name):
|
9 |
+
print(type(img_bgr))
|
10 |
+
palette = [
|
11 |
+
['background', [0, 0, 0]],
|
12 |
+
['red', [255, 0, 0]]
|
13 |
+
]
|
14 |
+
|
15 |
+
palette_dict = {}
|
16 |
+
for idx, each in enumerate(palette):
|
17 |
+
palette_dict[idx] = each[1]
|
18 |
+
|
19 |
+
if model_name == 'Mask2Former':
|
20 |
+
config_file = 'CVRP_configs/CVRP_mask2former.py'
|
21 |
+
checkpoint_file = 'checkpoint/Mask2Former.pth'
|
22 |
+
elif model_name == 'KNet':
|
23 |
+
config_file = 'CVRP_configs/CVRP_knet.py'
|
24 |
+
checkpoint_file = 'checkpoint/KNet.pth'
|
25 |
+
elif model_name == 'DeepLabV3+':
|
26 |
+
config_file = 'CVRP_configs/CVRP_deeplabv3plus.py'
|
27 |
+
checkpoint_file = 'checkpoint/DeepLabV3plus.pth'
|
28 |
+
elif model_name == 'Segformer':
|
29 |
+
config_file = 'CVRP_configs/CVRP_segformer.py'
|
30 |
+
checkpoint_file = 'checkpoint/Segformer.pth'
|
31 |
+
else:
|
32 |
+
return None, None
|
33 |
+
|
34 |
+
device = 'cuda:0'
|
35 |
+
|
36 |
+
model = init_model(config_file, checkpoint_file, device=device)
|
37 |
+
|
38 |
+
result = inference_model(model, img_bgr)
|
39 |
+
pred_mask = result.pred_sem_seg.data[0].cpu().numpy()
|
40 |
+
|
41 |
+
pred_mask_bgr = np.zeros((pred_mask.shape[0], pred_mask.shape[1], 3))
|
42 |
+
for idx in palette_dict.keys():
|
43 |
+
pred_mask_bgr[np.where(pred_mask == idx)] = palette_dict[idx]
|
44 |
+
pred_mask_bgr = pred_mask_bgr.astype('uint8')
|
45 |
+
|
46 |
+
pred_viz = cv2.addWeighted(img_bgr, 1, pred_mask_bgr, 1, 0)
|
47 |
+
|
48 |
+
torch.cuda.empty_cache()
|
49 |
+
|
50 |
+
return pred_viz, pred_mask_bgr
|
51 |
+
def run_segmentation(image_input, model_select):
|
52 |
+
if model_select not in ["Mask2Former", "KNet", "DeepLabV3+", "Segformer"]:
|
53 |
+
return None, None, [("No implementa", "Error"), ("", "")]
|
54 |
+
else:
|
55 |
+
color_img, binary_img = process_single_img(image_input, model_select)
|
56 |
+
return color_img, binary_img, [("", ""), ("Segmentation Finished", "normal")]
|
57 |
+
|
58 |
+
title = """<p><h1 align="center">CVRP</h1></p>"""
|
59 |
+
# 设置SAM参数
|
60 |
+
|
61 |
+
with gr.Blocks() as iface:
|
62 |
+
gr.Markdown(title)
|
63 |
+
with gr.Row():
|
64 |
+
with gr.Column():
|
65 |
+
image_input = gr.Image(interactive=True, visible=True, label="Input Image", height=360)
|
66 |
+
with gr.Row():
|
67 |
+
model_select = gr.Dropdown(choices=["Mask2Former", "KNet", "DeepLabV3+", "Segformer"], value="Mask2Former", label="Select model", visible=True)
|
68 |
+
run_button = gr.Button(value="Run", interactive=True, visible=True)
|
69 |
+
with gr.Row():
|
70 |
+
gr.Examples(
|
71 |
+
examples=[['assets/T42_1220.jpg', 'Mask2Former'], ['assets/02604.jpg', 'Mask2Former'], ['assets/T92_323.jpg', 'Mask2Former']],
|
72 |
+
inputs=[image_input, model_select])
|
73 |
+
|
74 |
+
with gr.Column():
|
75 |
+
color_output = gr.Image(interactive=False, visible=True, label="Color Image", height=360)
|
76 |
+
binary_output = gr.Image(interactive=False, visible=True, label="Binary Image", height=360)
|
77 |
+
run_status = gr.HighlightedText(
|
78 |
+
value=[("Text", "Error"), ("to be", "Label 2"), ("highlighted", "Label 3")], visible=True)
|
79 |
+
|
80 |
+
run_button.click(
|
81 |
+
fn=run_segmentation,
|
82 |
+
inputs=[image_input, model_select],
|
83 |
+
outputs=[color_output, binary_output, run_status]
|
84 |
+
)
|
85 |
+
|
86 |
+
|
87 |
+
iface.launch(debug=True, server_port=6006, server_name="127.0.0.1")
|