Update app.py
Browse files
app.py
CHANGED
@@ -1,111 +1,75 @@
|
|
1 |
import streamlit as st
|
2 |
-
from transformers import pipeline
|
3 |
from PIL import Image, ImageDraw
|
4 |
import torch
|
5 |
|
|
|
6 |
st.set_page_config(
|
7 |
-
page_title="Fraktur Detektion",
|
8 |
layout="wide",
|
|
|
9 |
initial_sidebar_state="collapsed"
|
10 |
)
|
11 |
|
12 |
-
#
|
13 |
st.markdown("""
|
14 |
<style>
|
|
|
|
|
|
|
|
|
|
|
|
|
15 |
.stApp {
|
16 |
-
|
17 |
padding: 0 !important;
|
|
|
18 |
}
|
19 |
|
20 |
-
|
21 |
-
--background-color: #ffffff;
|
22 |
-
--text-color: #1f2937;
|
23 |
-
--border-color: #e5e7eb;
|
24 |
-
--button-color: #2563eb;
|
25 |
-
--button-hover: #1d4ed8;
|
26 |
-
}
|
27 |
-
|
28 |
-
[data-theme="dark"] {
|
29 |
-
--background-color: #1f2937;
|
30 |
-
--text-color: #f3f4f6;
|
31 |
-
--border-color: #4b5563;
|
32 |
-
--button-color: #3b82f6;
|
33 |
-
--button-hover: #2563eb;
|
34 |
-
}
|
35 |
-
|
36 |
.stButton > button {
|
37 |
-
width: 100
|
38 |
-
background-color:
|
39 |
-
color: white
|
40 |
-
|
41 |
-
|
42 |
-
border
|
43 |
-
font-weight: 500
|
44 |
-
margin: 1rem 0 !important;
|
45 |
-
cursor: pointer !important;
|
46 |
}
|
47 |
|
48 |
.stButton > button:hover {
|
49 |
-
background-color:
|
50 |
}
|
51 |
|
|
|
52 |
.block-container {
|
53 |
-
padding:
|
54 |
max-width: 100% !important;
|
55 |
}
|
56 |
|
|
|
57 |
.stImage > img {
|
58 |
-
max-height:
|
59 |
-
width: auto
|
60 |
-
margin: 0 auto
|
61 |
-
}
|
62 |
-
|
63 |
-
.result-box {
|
64 |
-
padding: 0.375rem;
|
65 |
-
border-radius: 0.375rem;
|
66 |
-
margin: 0.25rem 0;
|
67 |
-
background: var(--background-color);
|
68 |
-
border: 1px solid var(--border-color);
|
69 |
-
color: var(--text-color);
|
70 |
-
}
|
71 |
-
|
72 |
-
h2, h3, h4 {
|
73 |
-
margin: 0.5rem 0 !important;
|
74 |
-
color: var(--text-color) !important;
|
75 |
-
font-size: 1rem !important;
|
76 |
-
}
|
77 |
-
|
78 |
-
#MainMenu, footer, header {
|
79 |
-
display: none !important;
|
80 |
}
|
81 |
|
82 |
-
|
83 |
-
|
|
|
|
|
84 |
border-radius: 0.375rem;
|
85 |
-
padding: 0.25rem;
|
86 |
-
}
|
87 |
-
|
88 |
-
div[data-testid="stFileUploader"] {
|
89 |
-
width: 100%;
|
90 |
-
}
|
91 |
-
|
92 |
-
/* Cache le message d'upload par défaut */
|
93 |
-
.uploadedFile small {
|
94 |
-
display: none !important;
|
95 |
}
|
96 |
</style>
|
97 |
""", unsafe_allow_html=True)
|
98 |
|
|
|
99 |
@st.cache_resource
|
100 |
def load_models():
|
101 |
return {
|
102 |
"KnochenAuge": pipeline("object-detection", model="D3STRON/bone-fracture-detr"),
|
103 |
-
"KnochenWächter": pipeline("image-classification",
|
104 |
-
model="Heem2/bone-fracture-detection-using-xray",
|
105 |
-
image_processor=AutoImageProcessor.from_pretrained("Heem2/bone-fracture-detection-using-xray")),
|
106 |
"RöntgenMeister": pipeline("image-classification",
|
107 |
-
model="nandodeomkar/autotrain-fracture-detection-using-google-vit-base-patch-16-54382127388"
|
108 |
-
image_processor=AutoImageProcessor.from_pretrained("nandodeomkar/autotrain-fracture-detection-using-google-vit-base-patch-16-54382127388"))
|
109 |
}
|
110 |
|
111 |
def draw_boxes(image, predictions):
|
@@ -113,98 +77,73 @@ def draw_boxes(image, predictions):
|
|
113 |
for pred in predictions:
|
114 |
if pred['label'].lower() == 'fracture' and pred['score'] > 0.6:
|
115 |
box = pred['box']
|
116 |
-
label = f"Fraktur ({pred['score']:.
|
117 |
-
color = "#
|
118 |
|
|
|
119 |
draw.rectangle(
|
120 |
[(box['xmin'], box['ymin']), (box['xmax'], box['ymax'])],
|
121 |
outline=color,
|
122 |
width=2
|
123 |
)
|
124 |
|
|
|
125 |
text_bbox = draw.textbbox((box['xmin'], box['ymin']-15), label)
|
126 |
draw.rectangle(text_bbox, fill=color)
|
127 |
draw.text((box['xmin'], box['ymin']-15), label, fill="white")
|
|
|
128 |
return image
|
129 |
|
130 |
def main():
|
131 |
-
#
|
132 |
-
|
133 |
-
|
134 |
-
|
135 |
-
|
136 |
-
|
137 |
-
|
138 |
-
|
139 |
-
|
140 |
-
|
141 |
-
|
142 |
-
|
143 |
-
|
144 |
-
|
145 |
-
|
146 |
-
|
147 |
-
|
148 |
-
# Bouton d'analyse dans la deuxième colonne
|
149 |
-
with col_button:
|
150 |
-
analyze_clicked = st.button("📋 Analysieren", key="analyze_btn", disabled=not uploaded_files)
|
151 |
-
|
152 |
-
# Analyse des images si le bouton est cliqué
|
153 |
-
if uploaded_files and analyze_clicked:
|
154 |
-
col1, col2 = st.columns([1, 1])
|
155 |
-
|
156 |
-
for idx, uploaded_file in enumerate(uploaded_files):
|
157 |
-
image = Image.open(uploaded_file)
|
158 |
-
|
159 |
-
# Analyse avec KnochenAuge (localisierung)
|
160 |
-
predictions = st.session_state.models["KnochenAuge"](image)
|
161 |
-
fractures_found = any(p['label'].lower() == 'fracture' and p['score'] > 0.6 for p in predictions)
|
162 |
-
|
163 |
-
# Afficher uniquement si des fractures sont détectées
|
164 |
-
if fractures_found:
|
165 |
-
with col1 if idx % 2 == 0 else col2:
|
166 |
-
result_image = image.copy()
|
167 |
-
result_image = draw_boxes(result_image, predictions)
|
168 |
-
st.image(result_image, caption=f"Bild {idx + 1}", use_column_width=True)
|
169 |
-
|
170 |
-
# Analyse KnochenWächter et RöntgenMeister
|
171 |
-
pred_wachter = st.session_state.models["KnochenWächter"](image)[0]
|
172 |
-
pred_meister = st.session_state.models["RöntgenMeister"](image)[0]
|
173 |
-
|
174 |
-
if pred_wachter['score'] > 0.6 or pred_meister['score'] > 0.6:
|
175 |
-
st.markdown(f"""
|
176 |
-
<div class='result-box'>
|
177 |
-
<span style='color: #2563eb'>KnochenWächter:</span> {pred_wachter['score']:.1%}<br>
|
178 |
-
<span style='color: #2563eb'>RöntgenMeister:</span> {pred_meister['score']:.1%}
|
179 |
-
</div>
|
180 |
-
""", unsafe_allow_html=True)
|
181 |
-
|
182 |
-
# Script pour la synchronisation du thème
|
183 |
-
st.markdown("""
|
184 |
-
<script>
|
185 |
-
function updateTheme(isDark) {
|
186 |
-
document.documentElement.setAttribute('data-theme', isDark ? 'dark' : 'light');
|
187 |
-
const root = document.documentElement;
|
188 |
-
if (isDark) {
|
189 |
-
root.style.setProperty('--background-color', '#1f2937');
|
190 |
-
root.style.setProperty('--text-color', '#f3f4f6');
|
191 |
-
root.style.setProperty('--border-color', '#4b5563');
|
192 |
-
} else {
|
193 |
-
root.style.setProperty('--background-color', '#ffffff');
|
194 |
-
root.style.setProperty('--text-color', '#1f2937');
|
195 |
-
root.style.setProperty('--border-color', '#e5e7eb');
|
196 |
-
}
|
197 |
-
}
|
198 |
-
|
199 |
-
window.addEventListener('message', function(e) {
|
200 |
-
if (e.data.type === 'theme-change') {
|
201 |
-
updateTheme(e.data.theme === 'dark');
|
202 |
-
}
|
203 |
-
});
|
204 |
|
205 |
-
|
206 |
-
|
207 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
208 |
|
209 |
if __name__ == "__main__":
|
210 |
main()
|
|
|
1 |
import streamlit as st
|
2 |
+
from transformers import pipeline
|
3 |
from PIL import Image, ImageDraw
|
4 |
import torch
|
5 |
|
6 |
+
# Configuration de la page
|
7 |
st.set_page_config(
|
|
|
8 |
layout="wide",
|
9 |
+
page_title="Fraktur Detektion",
|
10 |
initial_sidebar_state="collapsed"
|
11 |
)
|
12 |
|
13 |
+
# Style personnalisé
|
14 |
st.markdown("""
|
15 |
<style>
|
16 |
+
/* Cacher les éléments Streamlit par défaut */
|
17 |
+
#MainMenu {visibility: hidden;}
|
18 |
+
footer {visibility: hidden;}
|
19 |
+
header {visibility: hidden;}
|
20 |
+
|
21 |
+
/* Style personnalisé pour la page */
|
22 |
.stApp {
|
23 |
+
margin: 0;
|
24 |
padding: 0 !important;
|
25 |
+
max-width: 100%;
|
26 |
}
|
27 |
|
28 |
+
/* Style pour les boutons */
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
29 |
.stButton > button {
|
30 |
+
width: 100%;
|
31 |
+
background-color: #3b82f6;
|
32 |
+
color: white;
|
33 |
+
padding: 0.5rem 1rem;
|
34 |
+
border-radius: 0.375rem;
|
35 |
+
border: none;
|
36 |
+
font-weight: 500;
|
|
|
|
|
37 |
}
|
38 |
|
39 |
.stButton > button:hover {
|
40 |
+
background-color: #2563eb;
|
41 |
}
|
42 |
|
43 |
+
/* Container principal */
|
44 |
.block-container {
|
45 |
+
padding: 1rem !important;
|
46 |
max-width: 100% !important;
|
47 |
}
|
48 |
|
49 |
+
/* Style pour les images */
|
50 |
.stImage > img {
|
51 |
+
max-height: 300px;
|
52 |
+
width: auto;
|
53 |
+
margin: 0 auto;
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
54 |
}
|
55 |
|
56 |
+
/* Style pour les messages d'erreur */
|
57 |
+
.stAlert {
|
58 |
+
padding: 1rem;
|
59 |
+
margin: 1rem 0;
|
60 |
border-radius: 0.375rem;
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
61 |
}
|
62 |
</style>
|
63 |
""", unsafe_allow_html=True)
|
64 |
|
65 |
+
# Cache des modèles
|
66 |
@st.cache_resource
|
67 |
def load_models():
|
68 |
return {
|
69 |
"KnochenAuge": pipeline("object-detection", model="D3STRON/bone-fracture-detr"),
|
70 |
+
"KnochenWächter": pipeline("image-classification", model="Heem2/bone-fracture-detection-using-xray"),
|
|
|
|
|
71 |
"RöntgenMeister": pipeline("image-classification",
|
72 |
+
model="nandodeomkar/autotrain-fracture-detection-using-google-vit-base-patch-16-54382127388")
|
|
|
73 |
}
|
74 |
|
75 |
def draw_boxes(image, predictions):
|
|
|
77 |
for pred in predictions:
|
78 |
if pred['label'].lower() == 'fracture' and pred['score'] > 0.6:
|
79 |
box = pred['box']
|
80 |
+
label = f"Fraktur ({pred['score']:.1%})"
|
81 |
+
color = "#EF4444" # Rouge
|
82 |
|
83 |
+
# Dessiner le rectangle
|
84 |
draw.rectangle(
|
85 |
[(box['xmin'], box['ymin']), (box['xmax'], box['ymax'])],
|
86 |
outline=color,
|
87 |
width=2
|
88 |
)
|
89 |
|
90 |
+
# Ajouter le label
|
91 |
text_bbox = draw.textbbox((box['xmin'], box['ymin']-15), label)
|
92 |
draw.rectangle(text_bbox, fill=color)
|
93 |
draw.text((box['xmin'], box['ymin']-15), label, fill="white")
|
94 |
+
|
95 |
return image
|
96 |
|
97 |
def main():
|
98 |
+
# Chargement des modèles
|
99 |
+
models = load_models()
|
100 |
+
|
101 |
+
# Upload des images
|
102 |
+
uploaded_files = st.file_uploader(
|
103 |
+
"Röntgenbilder hochladen",
|
104 |
+
type=['png', 'jpg', 'jpeg'],
|
105 |
+
accept_multiple_files=True,
|
106 |
+
label_visibility="collapsed"
|
107 |
+
)
|
108 |
+
|
109 |
+
if uploaded_files:
|
110 |
+
# Bouton d'analyse
|
111 |
+
if st.button("Bilder analysieren", key="analyze_button"):
|
112 |
+
# Création des colonnes pour l'affichage
|
113 |
+
cols = st.columns(2)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
114 |
|
115 |
+
for idx, uploaded_file in enumerate(uploaded_files):
|
116 |
+
image = Image.open(uploaded_file)
|
117 |
+
|
118 |
+
# Analyse avec KnochenAuge
|
119 |
+
predictions = models["KnochenAuge"](image)
|
120 |
+
fractures_found = any(p['label'].lower() == 'fracture' and p['score'] > 0.6 for p in predictions)
|
121 |
+
|
122 |
+
if fractures_found:
|
123 |
+
with cols[idx % 2]:
|
124 |
+
# Créer une copie de l'image pour le dessin
|
125 |
+
result_image = image.copy()
|
126 |
+
result_image = draw_boxes(result_image, predictions)
|
127 |
+
st.image(result_image, use_column_width=True)
|
128 |
+
|
129 |
+
# Analyses supplémentaires
|
130 |
+
pred_wachter = models["KnochenWächter"](image)[0]
|
131 |
+
pred_meister = models["RöntgenMeister"](image)[0]
|
132 |
+
|
133 |
+
if pred_wachter['score'] > 0.6 or pred_meister['score'] > 0.6:
|
134 |
+
st.markdown(
|
135 |
+
f"""
|
136 |
+
<div style='background-color: #1F2937; color: white; padding: 1rem; border-radius: 0.375rem;'>
|
137 |
+
<div style='margin-bottom: 0.5rem;'>
|
138 |
+
<span style='color: #60A5FA;'>KnochenWächter:</span> {pred_wachter['score']:.1%}
|
139 |
+
</div>
|
140 |
+
<div>
|
141 |
+
<span style='color: #60A5FA;'>RöntgenMeister:</span> {pred_meister['score']:.1%}
|
142 |
+
</div>
|
143 |
+
</div>
|
144 |
+
""",
|
145 |
+
unsafe_allow_html=True
|
146 |
+
)
|
147 |
|
148 |
if __name__ == "__main__":
|
149 |
main()
|