Added bug fixes to the demo
Browse files- demo/src/compute.py +2 -2
- demo/src/gui.py +50 -38
demo/src/compute.py
CHANGED
@@ -1,6 +1,6 @@
|
|
1 |
import subprocess as sp
|
2 |
|
3 |
|
4 |
-
def run_model(fixed_path, moving_path, output_path):
|
5 |
sp.check_call(["ddmr", "--fixed", fixed_path, "--moving", moving_path, \
|
6 |
-
"-o", output_path, "-a",
|
|
|
1 |
import subprocess as sp
|
2 |
|
3 |
|
4 |
+
def run_model(fixed_path, moving_path, output_path, task):
|
5 |
sp.check_call(["ddmr", "--fixed", fixed_path, "--moving", moving_path, \
|
6 |
+
"-o", output_path, "-a", task, "--model", "BL-NS", "--original-resolution"])
|
demo/src/gui.py
CHANGED
@@ -28,8 +28,8 @@ class WebUI:
|
|
28 |
self.share = share
|
29 |
|
30 |
self.class_names = {
|
31 |
-
"
|
32 |
-
"
|
33 |
}
|
34 |
|
35 |
# define widgets not to be rendered immediantly, but later on
|
@@ -40,42 +40,49 @@ class WebUI:
|
|
40 |
step=1,
|
41 |
label="Which 2D slice to show",
|
42 |
)
|
43 |
-
self.volume_renderer = gr.Model3D(
|
44 |
-
clear_color=[0.0, 0.0, 0.0, 0.0],
|
45 |
-
label="3D Model",
|
46 |
-
visible=True,
|
47 |
-
elem_id="model-3d",
|
48 |
-
).style(height=512)
|
49 |
|
50 |
def set_class_name(self, value):
|
51 |
print("Changed task to:", value)
|
52 |
self.class_name = value
|
53 |
|
54 |
-
def combine_ct_and_seg(self, img, pred):
|
55 |
-
return (img, [(pred, self.class_name)])
|
56 |
-
|
57 |
def upload_file(self, file):
|
58 |
return file.name
|
59 |
|
60 |
-
def process(self,
|
61 |
-
|
62 |
-
|
63 |
-
path,
|
64 |
-
model_path=os.path.join(self.cwd, "resources/models/"),
|
65 |
-
task=self.class_names[self.class_name],
|
66 |
-
name=self.result_names[self.class_name],
|
67 |
-
)
|
68 |
-
nifti_to_glb("prediction.nii.gz")
|
69 |
|
70 |
-
self.
|
71 |
-
self.pred_images = load_pred_volume_to_numpy("./prediction.nii.gz")
|
72 |
-
return "./prediction.obj"
|
73 |
|
74 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
75 |
k = int(k) - 1
|
76 |
-
out = [gr.
|
77 |
-
out[k] = gr.
|
78 |
-
self.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
79 |
visible=True,
|
80 |
)
|
81 |
return out
|
@@ -95,13 +102,13 @@ class WebUI:
|
|
95 |
"""
|
96 |
with gr.Blocks(css=css) as demo:
|
97 |
with gr.Row():
|
98 |
-
file_output = gr.File(file_count="
|
99 |
file_output.upload(self.upload_file, file_output, file_output)
|
100 |
|
101 |
model_selector = gr.Dropdown(
|
102 |
list(self.class_names.keys()),
|
103 |
label="Task",
|
104 |
-
info="Which task to perform registration
|
105 |
multiselect=False,
|
106 |
size="sm",
|
107 |
)
|
@@ -117,7 +124,7 @@ class WebUI:
|
|
117 |
run_btn.click(
|
118 |
fn=lambda x: self.process(x),
|
119 |
inputs=file_output,
|
120 |
-
outputs=
|
121 |
)
|
122 |
|
123 |
with gr.Row():
|
@@ -135,27 +142,32 @@ class WebUI:
|
|
135 |
with gr.Row():
|
136 |
with gr.Box():
|
137 |
with gr.Column():
|
138 |
-
|
139 |
for i in range(self.nb_slider_items):
|
140 |
visibility = True if i == 1 else False
|
141 |
-
t = gr.
|
142 |
visible=visibility, elem_id="model-2d"
|
143 |
).style(
|
144 |
-
color_map={self.class_name: "#ffae00"},
|
145 |
height=512,
|
146 |
width=512,
|
147 |
)
|
148 |
-
|
|
|
|
|
|
|
149 |
|
150 |
self.slider.input(
|
151 |
-
self.
|
|
|
|
|
|
|
|
|
|
|
|
|
152 |
)
|
153 |
|
154 |
self.slider.render()
|
155 |
|
156 |
-
with gr.Box():
|
157 |
-
self.volume_renderer.render()
|
158 |
-
|
159 |
# sharing app publicly -> share=True:
|
160 |
# https://gradio.app/sharing-your-app/
|
161 |
# inference times > 60 seconds -> need queue():
|
|
|
28 |
self.share = share
|
29 |
|
30 |
self.class_names = {
|
31 |
+
"Brain": "B",
|
32 |
+
"Liver": "L"
|
33 |
}
|
34 |
|
35 |
# define widgets not to be rendered immediantly, but later on
|
|
|
40 |
step=1,
|
41 |
label="Which 2D slice to show",
|
42 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
43 |
|
44 |
def set_class_name(self, value):
|
45 |
print("Changed task to:", value)
|
46 |
self.class_name = value
|
47 |
|
|
|
|
|
|
|
48 |
def upload_file(self, file):
|
49 |
return file.name
|
50 |
|
51 |
+
def process(self, mesh_file_names):
|
52 |
+
fixed_image_path = mesh_file_names[0].name
|
53 |
+
moving_image_path = mesh_file_names[1].name
|
|
|
|
|
|
|
|
|
|
|
|
|
54 |
|
55 |
+
run_model(fixed_path, moving_path, output_path, self.class_names[self.class_name])
|
|
|
|
|
56 |
|
57 |
+
self.fixed_images = load_ct_to_numpy(fixed_image_path)
|
58 |
+
self.moving_images = load_ct_to_numpy(moving_image_path)
|
59 |
+
#self.pred_images = load_ct_to_numpy("./prediction.nii.gz")
|
60 |
+
self.pred_images = np.ones_like(moving_images)
|
61 |
+
return None
|
62 |
+
|
63 |
+
def get_fixed_image(self, k):
|
64 |
+
k = int(k) - 1
|
65 |
+
out = [gr.Image.update(visible=False)] * self.nb_slider_items
|
66 |
+
out[k] = gr.Image.update(
|
67 |
+
self.fixed_images[k]
|
68 |
+
visible=True,
|
69 |
+
)
|
70 |
+
return out
|
71 |
+
|
72 |
+
def get_moving_image(self, k):
|
73 |
k = int(k) - 1
|
74 |
+
out = [gr.Image.update(visible=False)] * self.nb_slider_items
|
75 |
+
out[k] = gr.Image.update(
|
76 |
+
self.moving_images[k]
|
77 |
+
visible=True,
|
78 |
+
)
|
79 |
+
return out
|
80 |
+
|
81 |
+
def get_pred_image(self, k):
|
82 |
+
k = int(k) - 1
|
83 |
+
out = [gr.Image.update(visible=False)] * self.nb_slider_items
|
84 |
+
out[k] = gr.Image.update(
|
85 |
+
self.pred_images[k]
|
86 |
visible=True,
|
87 |
)
|
88 |
return out
|
|
|
102 |
"""
|
103 |
with gr.Blocks(css=css) as demo:
|
104 |
with gr.Row():
|
105 |
+
file_output = gr.File(file_count="multiple", elem_id="upload")
|
106 |
file_output.upload(self.upload_file, file_output, file_output)
|
107 |
|
108 |
model_selector = gr.Dropdown(
|
109 |
list(self.class_names.keys()),
|
110 |
label="Task",
|
111 |
+
info="Which task to perform image-to-registration on",
|
112 |
multiselect=False,
|
113 |
size="sm",
|
114 |
)
|
|
|
124 |
run_btn.click(
|
125 |
fn=lambda x: self.process(x),
|
126 |
inputs=file_output,
|
127 |
+
outputs=None,
|
128 |
)
|
129 |
|
130 |
with gr.Row():
|
|
|
142 |
with gr.Row():
|
143 |
with gr.Box():
|
144 |
with gr.Column():
|
145 |
+
fixed_images = []
|
146 |
for i in range(self.nb_slider_items):
|
147 |
visibility = True if i == 1 else False
|
148 |
+
t = gr.Image(
|
149 |
visible=visibility, elem_id="model-2d"
|
150 |
).style(
|
|
|
151 |
height=512,
|
152 |
width=512,
|
153 |
)
|
154 |
+
fixed_images.append(t)
|
155 |
+
|
156 |
+
moving_images = fixed_images.copy()
|
157 |
+
pred_images = fixed_images.copy()
|
158 |
|
159 |
self.slider.input(
|
160 |
+
self.get_fixed_image, self.slider, fixed_images
|
161 |
+
)
|
162 |
+
self.slider.input(
|
163 |
+
self.get_moving_image, self.slider, moving_images
|
164 |
+
)
|
165 |
+
self.slider.input(
|
166 |
+
self.get_pred_image, self.slider, pred_images
|
167 |
)
|
168 |
|
169 |
self.slider.render()
|
170 |
|
|
|
|
|
|
|
171 |
# sharing app publicly -> share=True:
|
172 |
# https://gradio.app/sharing-your-app/
|
173 |
# inference times > 60 seconds -> need queue():
|