Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,24 +1,46 @@
|
|
| 1 |
# -*- coding: utf-8 -*-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
import torch
|
| 3 |
import torch.nn as nn
|
| 4 |
import torch.nn.functional as F
|
| 5 |
-
|
| 6 |
-
import numpy as np
|
| 7 |
import gradio as gr
|
| 8 |
-
import gdown
|
| 9 |
-
import os
|
| 10 |
-
from PIL import Image
|
| 11 |
|
| 12 |
-
#
|
| 13 |
-
#
|
| 14 |
-
#
|
| 15 |
DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
| 16 |
-
IMG_SIZE = 512
|
| 17 |
-
|
| 18 |
-
|
| 19 |
-
|
| 20 |
-
#
|
| 21 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 22 |
class ConvBlock(nn.Module):
|
| 23 |
def __init__(self, in_ch, out_ch):
|
| 24 |
super().__init__()
|
|
@@ -28,23 +50,18 @@ class ConvBlock(nn.Module):
|
|
| 28 |
nn.ReLU(inplace=True),
|
| 29 |
nn.Conv2d(out_ch, out_ch, 3, padding=1),
|
| 30 |
nn.BatchNorm2d(out_ch),
|
| 31 |
-
nn.ReLU(inplace=True)
|
| 32 |
)
|
| 33 |
-
|
| 34 |
def forward(self, x):
|
| 35 |
return self.block(x)
|
| 36 |
|
| 37 |
class DFUTissueSegNet(nn.Module):
|
| 38 |
def __init__(self, num_classes=3):
|
| 39 |
super().__init__()
|
| 40 |
-
self.
|
| 41 |
-
self.
|
| 42 |
-
self.
|
| 43 |
-
self.
|
| 44 |
-
self.encoder3 = ConvBlock(128, 256)
|
| 45 |
-
self.pool3 = nn.MaxPool2d(2)
|
| 46 |
-
self.encoder4 = ConvBlock(256, 512)
|
| 47 |
-
self.pool4 = nn.MaxPool2d(2)
|
| 48 |
|
| 49 |
self.center = ConvBlock(512, 1024)
|
| 50 |
|
|
@@ -60,43 +77,29 @@ class DFUTissueSegNet(nn.Module):
|
|
| 60 |
self.final = nn.Conv2d(64, num_classes, 1)
|
| 61 |
|
| 62 |
def forward(self, x):
|
| 63 |
-
e1 = self.
|
| 64 |
-
e2 = self.
|
| 65 |
-
e3 = self.
|
| 66 |
-
e4 = self.
|
| 67 |
-
c
|
| 68 |
-
|
| 69 |
-
d4 = self.up4(c)
|
| 70 |
-
d4 = torch.cat([
|
| 71 |
-
|
| 72 |
-
|
| 73 |
-
|
| 74 |
-
|
| 75 |
-
d3 = self.dec3(d3)
|
| 76 |
-
|
| 77 |
-
d2 = self.up2(d3)
|
| 78 |
-
d2 = torch.cat([d2, e2], dim=1)
|
| 79 |
-
d2 = self.dec2(d2)
|
| 80 |
-
|
| 81 |
-
d1 = self.up1(d2)
|
| 82 |
-
d1 = torch.cat([d1, e1], dim=1)
|
| 83 |
-
d1 = self.dec1(d1)
|
| 84 |
-
|
| 85 |
return torch.sigmoid(self.final(d1))
|
| 86 |
|
| 87 |
-
# =========================================================
|
| 88 |
-
# تحميل النموذج من Google Drive
|
| 89 |
-
# =========================================================
|
| 90 |
segmenter = None
|
| 91 |
|
|
|
|
|
|
|
|
|
|
| 92 |
def initialize_model():
|
| 93 |
-
"""تحميل DFUTissueSegNet من Google Drive مع دعم checkpoint"""
|
| 94 |
global segmenter
|
| 95 |
-
MODEL_URL = "https://drive.google.com/uc?id=1Ovaczsjdp3E-_gYF2pbUibDjPWAC1a6c"
|
| 96 |
-
MODEL_PATH = "best_model_5.pth"
|
| 97 |
-
|
| 98 |
if not os.path.exists(MODEL_PATH):
|
| 99 |
-
print("📥 تحميل
|
| 100 |
gdown.download(MODEL_URL, MODEL_PATH, quiet=False)
|
| 101 |
|
| 102 |
try:
|
|
@@ -104,12 +107,13 @@ def initialize_model():
|
|
| 104 |
segmenter = DFUTissueSegNet(num_classes=3)
|
| 105 |
checkpoint = torch.load(MODEL_PATH, map_location=DEVICE)
|
| 106 |
|
| 107 |
-
#
|
| 108 |
-
if "state_dict" in checkpoint:
|
| 109 |
state_dict = checkpoint["state_dict"]
|
| 110 |
else:
|
| 111 |
state_dict = checkpoint
|
| 112 |
|
|
|
|
| 113 |
clean_state = {}
|
| 114 |
for k, v in state_dict.items():
|
| 115 |
nk = k.replace("module.", "").replace("model.", "")
|
|
@@ -118,111 +122,143 @@ def initialize_model():
|
|
| 118 |
segmenter.load_state_dict(clean_state, strict=False)
|
| 119 |
segmenter.to(DEVICE)
|
| 120 |
segmenter.eval()
|
| 121 |
-
print("✅
|
| 122 |
except Exception as e:
|
| 123 |
print(f"❌ فشل تحميل النموذج: {e}")
|
| 124 |
-
import traceback
|
| 125 |
-
traceback.print_exc()
|
| 126 |
segmenter = None
|
| 127 |
|
| 128 |
-
#
|
| 129 |
-
#
|
| 130 |
-
#
|
| 131 |
-
def
|
| 132 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 133 |
if segmenter is None:
|
| 134 |
-
return
|
| 135 |
-
|
| 136 |
-
|
| 137 |
-
|
| 138 |
-
|
| 139 |
-
|
| 140 |
-
|
| 141 |
-
|
| 142 |
-
|
| 143 |
-
|
| 144 |
-
|
| 145 |
-
|
| 146 |
-
|
| 147 |
-
|
| 148 |
-
|
| 149 |
-
|
| 150 |
-
|
| 151 |
-
|
| 152 |
-
|
| 153 |
-
|
| 154 |
-
|
| 155 |
-
|
| 156 |
-
|
| 157 |
-
|
| 158 |
-
|
| 159 |
-
|
| 160 |
-
|
| 161 |
-
|
| 162 |
-
|
| 163 |
-
|
| 164 |
-
|
| 165 |
-
|
| 166 |
-
|
| 167 |
-
|
| 168 |
-
|
| 169 |
-
|
| 170 |
-
|
| 171 |
-
|
| 172 |
-
|
| 173 |
-
|
| 174 |
-
|
| 175 |
-
|
| 176 |
-
|
| 177 |
-
|
| 178 |
-
|
| 179 |
-
|
| 180 |
-
|
| 181 |
-
|
| 182 |
-
|
| 183 |
-
|
| 184 |
-
|
| 185 |
-
|
| 186 |
-
|
| 187 |
-
|
| 188 |
-
|
| 189 |
-
|
| 190 |
-
|
| 191 |
-
|
| 192 |
-
|
| 193 |
-
|
| 194 |
-
|
| 195 |
-
|
| 196 |
-
|
| 197 |
-
|
| 198 |
-
|
| 199 |
-
|
| 200 |
-
|
| 201 |
-
|
| 202 |
-
|
| 203 |
-
#
|
| 204 |
# واجهة Gradio
|
| 205 |
-
#
|
| 206 |
-
|
| 207 |
-
|
| 208 |
-
|
| 209 |
-
|
| 210 |
-
|
| 211 |
-
|
| 212 |
-
|
| 213 |
-
|
| 214 |
-
|
| 215 |
-
|
| 216 |
-
|
| 217 |
-
|
| 218 |
-
|
| 219 |
-
|
| 220 |
-
|
| 221 |
-
|
| 222 |
-
|
| 223 |
-
|
| 224 |
-
|
|
|
|
|
|
|
|
|
|
| 225 |
|
|
|
|
|
|
|
|
|
|
| 226 |
if __name__ == "__main__":
|
| 227 |
-
print("
|
| 228 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
# -*- coding: utf-8 -*-
|
| 2 |
+
# -*- coding: utf-8 -*-
|
| 3 |
+
"""
|
| 4 |
+
واجهة موحّدة لتحليل صورة القدم باستخدام DFUTissueSegNet (من Google Drive)
|
| 5 |
+
- ألوان متعددة: قرحة=أحمر، Slough=أصفر، نخر=أسود
|
| 6 |
+
- حساب نسب كل نوع + مستوى خطورة إجمالي
|
| 7 |
+
- Legend داخل Gradio
|
| 8 |
+
"""
|
| 9 |
+
|
| 10 |
+
import os
|
| 11 |
+
import cv2
|
| 12 |
+
import gdown
|
| 13 |
+
import numpy as np
|
| 14 |
+
from PIL import Image
|
| 15 |
+
|
| 16 |
import torch
|
| 17 |
import torch.nn as nn
|
| 18 |
import torch.nn.functional as F
|
| 19 |
+
|
|
|
|
| 20 |
import gradio as gr
|
|
|
|
|
|
|
|
|
|
| 21 |
|
| 22 |
+
# ================================
|
| 23 |
+
# إعدادات عامة
|
| 24 |
+
# ================================
|
| 25 |
DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
| 26 |
+
IMG_SIZE = 512 # حجم الإدخال كما في تدريب DFUTissueSegNet
|
| 27 |
+
THRESHOLD = 0.4 # عتبة تحويل الاحتمال إلى قناع (جرّبي 0.35..0.6 حسب بياناتك)
|
| 28 |
+
MODEL_PATH = "best_model_5.pth"
|
| 29 |
+
|
| 30 |
+
# غيّري الـ File ID أدناه لملفّك على Google Drive عند الحاجة
|
| 31 |
+
# مثال: https://drive.google.com/file/d/FILE_ID/view => استخدمي: https://drive.google.com/uc?id=FILE_ID
|
| 32 |
+
MODEL_URL = "https://drive.google.com/uc?id=1Ovaczsjdp3E-_gYF2pbUibDjPWAC1a6c"
|
| 33 |
+
|
| 34 |
+
CLASS_NAMES = ["قرحة (Granulation)", "Slough (أنسجة ميتة جزئيًا)", "نخر (Necrotic)"]
|
| 35 |
+
CLASS_COLORS = {
|
| 36 |
+
"قرحة (Granulation)": (255, 0, 0), # أحمر
|
| 37 |
+
"Slough (أنسجة ميتة جزئيًا)": (255, 255, 0), # أصفر
|
| 38 |
+
"نخر (Necrotic)": (0, 0, 0) # أسود
|
| 39 |
+
}
|
| 40 |
+
|
| 41 |
+
# ================================
|
| 42 |
+
# نموذج DFUTissueSegNet
|
| 43 |
+
# ================================
|
| 44 |
class ConvBlock(nn.Module):
|
| 45 |
def __init__(self, in_ch, out_ch):
|
| 46 |
super().__init__()
|
|
|
|
| 50 |
nn.ReLU(inplace=True),
|
| 51 |
nn.Conv2d(out_ch, out_ch, 3, padding=1),
|
| 52 |
nn.BatchNorm2d(out_ch),
|
| 53 |
+
nn.ReLU(inplace=True),
|
| 54 |
)
|
|
|
|
| 55 |
def forward(self, x):
|
| 56 |
return self.block(x)
|
| 57 |
|
| 58 |
class DFUTissueSegNet(nn.Module):
|
| 59 |
def __init__(self, num_classes=3):
|
| 60 |
super().__init__()
|
| 61 |
+
self.enc1 = ConvBlock(3, 64); self.pool1 = nn.MaxPool2d(2)
|
| 62 |
+
self.enc2 = ConvBlock(64, 128); self.pool2 = nn.MaxPool2d(2)
|
| 63 |
+
self.enc3 = ConvBlock(128, 256);self.pool3 = nn.MaxPool2d(2)
|
| 64 |
+
self.enc4 = ConvBlock(256, 512);self.pool4 = nn.MaxPool2d(2)
|
|
|
|
|
|
|
|
|
|
|
|
|
| 65 |
|
| 66 |
self.center = ConvBlock(512, 1024)
|
| 67 |
|
|
|
|
| 77 |
self.final = nn.Conv2d(64, num_classes, 1)
|
| 78 |
|
| 79 |
def forward(self, x):
|
| 80 |
+
e1 = self.enc1(x)
|
| 81 |
+
e2 = self.enc2(self.pool1(e1))
|
| 82 |
+
e3 = self.enc3(self.pool2(e2))
|
| 83 |
+
e4 = self.enc4(self.pool3(e3))
|
| 84 |
+
c = self.center(self.pool4(e4))
|
| 85 |
+
|
| 86 |
+
d4 = self.up4(c); d4 = torch.cat([d4, e4], dim=1); d4 = self.dec4(d4)
|
| 87 |
+
d3 = self.up3(d4); d3 = torch.cat([d3, e3], dim=1); d3 = self.dec3(d3)
|
| 88 |
+
d2 = self.up2(d3); d2 = torch.cat([d2, e2], dim=1); d2 = self.dec2(d2)
|
| 89 |
+
d1 = self.up1(d2); d1 = torch.cat([d1, e1], dim=1); d1 = self.dec1(d1)
|
| 90 |
+
|
| 91 |
+
# مخرجات احتمالية 0..1 لكل قناة
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 92 |
return torch.sigmoid(self.final(d1))
|
| 93 |
|
|
|
|
|
|
|
|
|
|
| 94 |
segmenter = None
|
| 95 |
|
| 96 |
+
# ================================
|
| 97 |
+
# تحميل النموذج (من Google Drive)
|
| 98 |
+
# ================================
|
| 99 |
def initialize_model():
|
|
|
|
| 100 |
global segmenter
|
|
|
|
|
|
|
|
|
|
| 101 |
if not os.path.exists(MODEL_PATH):
|
| 102 |
+
print("📥 تحميل النموذج من Google Drive...")
|
| 103 |
gdown.download(MODEL_URL, MODEL_PATH, quiet=False)
|
| 104 |
|
| 105 |
try:
|
|
|
|
| 107 |
segmenter = DFUTissueSegNet(num_classes=3)
|
| 108 |
checkpoint = torch.load(MODEL_PATH, map_location=DEVICE)
|
| 109 |
|
| 110 |
+
# التعامل مع checkpoints التي تحتوي state_dict
|
| 111 |
+
if isinstance(checkpoint, dict) and "state_dict" in checkpoint:
|
| 112 |
state_dict = checkpoint["state_dict"]
|
| 113 |
else:
|
| 114 |
state_dict = checkpoint
|
| 115 |
|
| 116 |
+
# تنظيف البوادئ الشائعة
|
| 117 |
clean_state = {}
|
| 118 |
for k, v in state_dict.items():
|
| 119 |
nk = k.replace("module.", "").replace("model.", "")
|
|
|
|
| 122 |
segmenter.load_state_dict(clean_state, strict=False)
|
| 123 |
segmenter.to(DEVICE)
|
| 124 |
segmenter.eval()
|
| 125 |
+
print("✅ النموذج جاهز.")
|
| 126 |
except Exception as e:
|
| 127 |
print(f"❌ فشل تحميل النموذج: {e}")
|
| 128 |
+
import traceback; traceback.print_exc()
|
|
|
|
| 129 |
segmenter = None
|
| 130 |
|
| 131 |
+
# ================================
|
| 132 |
+
# أدوات مساعدة
|
| 133 |
+
# ================================
|
| 134 |
+
def ensure_rgb(np_img):
|
| 135 |
+
if np_img.ndim == 2:
|
| 136 |
+
return cv2.cvtColor(np_img, cv2.COLOR_GRAY2RGB)
|
| 137 |
+
if np_img.shape[-1] == 4:
|
| 138 |
+
return cv2.cvtColor(np_img, cv2.COLOR_RGBA2RGB)
|
| 139 |
+
return np_img
|
| 140 |
+
|
| 141 |
+
def apply_legend_markdown():
|
| 142 |
+
return """
|
| 143 |
+
### 🧭 مفتاح الألوان (Legend)
|
| 144 |
+
- 🩸 **أحمر** → نسيج قرحة (Granulation)
|
| 145 |
+
- 🟡 **أصفر** → نسيج ميت جزئيًا (Slough)
|
| 146 |
+
- ⚫ **أسود** → نسيج نخر (Necrotic)
|
| 147 |
+
"""
|
| 148 |
+
|
| 149 |
+
# ================================
|
| 150 |
+
# التجزئة + الحساب + التلوين
|
| 151 |
+
# ================================
|
| 152 |
+
def segment_and_color(pil_img: Image.Image):
|
| 153 |
+
"""
|
| 154 |
+
يُرجع:
|
| 155 |
+
- blended: الصورة مدموج عليها القناع اللوني
|
| 156 |
+
- mask_rgb: القناع اللوني (RGB)
|
| 157 |
+
- stats: نسب كل فئة + الإجمالي + مستوى الخطورة
|
| 158 |
+
"""
|
| 159 |
if segmenter is None:
|
| 160 |
+
return pil_img, pil_img, {"خطأ": "النموذج غير مهيأ"}
|
| 161 |
+
|
| 162 |
+
# 1) التحضير
|
| 163 |
+
img_np = ensure_rgb(np.array(pil_img))
|
| 164 |
+
h, w = img_np.shape[:2]
|
| 165 |
+
|
| 166 |
+
# 2) التحجيم + التطبيع كما في التدريب
|
| 167 |
+
img_resized = cv2.resize(img_np, (IMG_SIZE, IMG_SIZE))
|
| 168 |
+
img_norm = img_resized.astype(np.float32) / 255.0
|
| 169 |
+
img_norm = (img_norm - 0.5) / 0.5 # (x - 0.5) / 0.5
|
| 170 |
+
|
| 171 |
+
tensor = torch.from_numpy(img_norm).permute(2, 0, 1).unsqueeze(0).to(DEVICE)
|
| 172 |
+
|
| 173 |
+
# 3) التنبؤ
|
| 174 |
+
with torch.no_grad():
|
| 175 |
+
probs = segmenter(tensor).cpu().squeeze(0).numpy() # (3, H, W) احتمالات
|
| 176 |
+
|
| 177 |
+
# 4) أقنعة ثنائية لكل فئة
|
| 178 |
+
masks = (probs >= THRESHOLD).astype(np.uint8) # 0/1
|
| 179 |
+
# إزالة الضوضاء البسيطة
|
| 180 |
+
kernel = np.ones((5, 5), np.uint8)
|
| 181 |
+
for i in range(masks.shape[0]):
|
| 182 |
+
masks[i] = cv2.morphologyEx(masks[i], cv2.MORPH_OPEN, kernel)
|
| 183 |
+
masks[i] = cv2.morphologyEx(masks[i], cv2.MORPH_CLOSE, kernel)
|
| 184 |
+
|
| 185 |
+
# 5) حساب النسب على أبعاد الإدخال ثم إعادة القياس
|
| 186 |
+
total_pixels_input = IMG_SIZE * IMG_SIZE
|
| 187 |
+
ratios = {
|
| 188 |
+
CLASS_NAMES[0]: np.sum(masks[0]) / total_pixels_input * 100,
|
| 189 |
+
CLASS_NAMES[1]: np.sum(masks[1]) / total_pixels_input * 100,
|
| 190 |
+
CLASS_NAMES[2]: np.sum(masks[2]) / total_pixels_input * 100,
|
| 191 |
+
}
|
| 192 |
+
total_ratio = sum(ratios.values())
|
| 193 |
+
|
| 194 |
+
# 6) إنشاء قناع لوني على حجم الإدخال ثم إعادته لحجم الصورة الأصلي
|
| 195 |
+
color_mask = np.zeros((IMG_SIZE, IMG_SIZE, 3), dtype=np.uint8)
|
| 196 |
+
color_mask[masks[0] == 1] = CLASS_COLORS[CLASS_NAMES[0]]
|
| 197 |
+
color_mask[masks[1] == 1] = CLASS_COLORS[CLASS_NAMES[1]]
|
| 198 |
+
color_mask[masks[2] == 1] = CLASS_COLORS[CLASS_NAMES[2]]
|
| 199 |
+
|
| 200 |
+
mask_rgb = cv2.resize(color_mask, (w, h), interpolation=cv2.INTER_NEAREST)
|
| 201 |
+
|
| 202 |
+
# 7) دمج القناع مع الصورة (ألفا ~0.5)
|
| 203 |
+
alpha = (cv2.cvtColor(mask_rgb, cv2.COLOR_RGB2GRAY) > 0).astype(np.float32)[..., None] * 0.5
|
| 204 |
+
blended = (alpha * mask_rgb + (1 - alpha) * img_np).astype(np.uint8)
|
| 205 |
+
|
| 206 |
+
# 8) مستوى الخطورة
|
| 207 |
+
if total_ratio == 0:
|
| 208 |
+
risk = "No Risk 🟢"
|
| 209 |
+
elif total_ratio <= 1:
|
| 210 |
+
risk = "Low Risk 🟡"
|
| 211 |
+
elif total_ratio <= 5:
|
| 212 |
+
risk = "Medium Risk 🟠"
|
| 213 |
+
else:
|
| 214 |
+
risk = "High Risk 🔴"
|
| 215 |
+
|
| 216 |
+
stats = {
|
| 217 |
+
"نِسَب_الأنسجة": {
|
| 218 |
+
CLASS_NAMES[0]: f"{ratios[CLASS_NAMES[0]]:.2f}%",
|
| 219 |
+
CLASS_NAMES[1]: f"{ratios[CLASS_NAMES[1]]:.2f}%",
|
| 220 |
+
CLASS_NAMES[2]: f"{ratios[CLASS_NAMES[2]]:.2f}%",
|
| 221 |
+
"الإجمالي": f"{total_ratio:.2f}%",
|
| 222 |
+
},
|
| 223 |
+
"مستوى_الخطورة": risk,
|
| 224 |
+
"ملاحظات": "التحليل يعتمد على DFUTissueSegNet متعدد الفئات (حجم 512 وتطبيع (x-0.5)/0.5)."
|
| 225 |
+
}
|
| 226 |
+
|
| 227 |
+
return Image.fromarray(blended), Image.fromarray(mask_rgb), stats
|
| 228 |
+
|
| 229 |
+
# ================================
|
| 230 |
# واجهة Gradio
|
| 231 |
+
# ================================
|
| 232 |
+
def build_ui():
|
| 233 |
+
with gr.Blocks(title="تحليل قرحة القدم - DFUTissueSegNet", theme=gr.themes.Soft()) as demo:
|
| 234 |
+
gr.Markdown("# 🦶 نظام تحليل قرحة القدم السكري - صورة واحدة")
|
| 235 |
+
gr.Markdown("يعتمد على **DFUTissueSegNet** لتجزئة الأنسجة وحساب نسبها، مع تلوين واضح وLegend.")
|
| 236 |
+
|
| 237 |
+
with gr.Row():
|
| 238 |
+
with gr.Column(scale=1):
|
| 239 |
+
input_img = gr.Image(type="pil", label="📤 ارفع صورة القدم", height=320)
|
| 240 |
+
analyze_btn = gr.Button("🔍 بدء التحليل", variant="primary")
|
| 241 |
+
legend = gr.Markdown(apply_legend_markdown())
|
| 242 |
+
|
| 243 |
+
with gr.Column(scale=1):
|
| 244 |
+
out_blended = gr.Image(type="pil", label="🩸 الصورة مع القناع", height=320)
|
| 245 |
+
out_mask = gr.Image(type="pil", label="🧩 القناع اللوني", height=320)
|
| 246 |
+
out_json = gr.JSON(label="📊 التقرير التفصيلي")
|
| 247 |
+
|
| 248 |
+
analyze_btn.click(
|
| 249 |
+
fn=segment_and_color,
|
| 250 |
+
inputs=[input_img],
|
| 251 |
+
outputs=[out_blended, out_mask, out_json]
|
| 252 |
+
)
|
| 253 |
+
return demo
|
| 254 |
|
| 255 |
+
# ================================
|
| 256 |
+
# تشغيل التطبيق
|
| 257 |
+
# ================================
|
| 258 |
if __name__ == "__main__":
|
| 259 |
+
print("🚀 تهيئة النموذج...")
|
| 260 |
+
initialize_model()
|
| 261 |
+
app = build_ui()
|
| 262 |
+
# ملاحظة: على Spaces لا حاجة لـ share=True
|
| 263 |
+
app.launch(server_name="0.0.0.0", server_port=7860, share=False)
|
| 264 |
+
|