Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,35 +1,19 @@
|
|
| 1 |
-
from flask import Flask, session, request, redirect, url_for, render_template_string, send_file
|
| 2 |
-
import datetime
|
| 3 |
-
import os
|
| 4 |
-
import secrets
|
| 5 |
import torch
|
| 6 |
-
from PIL import Image, ImageDraw
|
| 7 |
from transformers import GroundingDinoProcessor
|
| 8 |
from modeling_grounding_dino import GroundingDinoForObjectDetection
|
|
|
|
|
|
|
| 9 |
from itertools import cycle
|
|
|
|
|
|
|
|
|
|
| 10 |
import tempfile
|
| 11 |
-
|
| 12 |
-
|
| 13 |
-
|
| 14 |
-
app.secret_key = os.environ.get('SECRET_KEY', secrets.token_hex(16))
|
| 15 |
-
SECRET_PASSWORD = "VeronaTrento25!"
|
| 16 |
-
app.permanent_session_lifetime = datetime.timedelta(hours=24)
|
| 17 |
-
|
| 18 |
-
# ===== AUTHENTICATION FUNCTIONS =====
|
| 19 |
-
def is_authenticated():
|
| 20 |
-
return session.get('authenticated', False)
|
| 21 |
-
|
| 22 |
-
def require_auth(f):
|
| 23 |
-
def decorated_function(*args, **kwargs):
|
| 24 |
-
if not is_authenticated():
|
| 25 |
-
return redirect(url_for('login'))
|
| 26 |
-
return f(*args, **kwargs)
|
| 27 |
-
decorated_function.__name__ = f.__name__
|
| 28 |
-
return decorated_function
|
| 29 |
-
|
| 30 |
-
# ===== ML MODEL SETUP =====
|
| 31 |
-
DEVICE = "cpu"
|
| 32 |
model_id = "fushh7/llmdet_swin_tiny_hf"
|
|
|
|
| 33 |
|
| 34 |
print(f"[INFO] Using device: {DEVICE}")
|
| 35 |
print(f"[INFO] Loading model from {model_id}...")
|
|
@@ -40,346 +24,266 @@ model.eval()
|
|
| 40 |
|
| 41 |
print("[INFO] Model loaded successfully.")
|
| 42 |
|
| 43 |
-
# Pre-defined palette
|
| 44 |
BOX_COLORS = [
|
| 45 |
"deepskyblue", "red", "lime", "dodgerblue",
|
| 46 |
-
"cyan", "magenta", "yellow",
|
|
|
|
| 47 |
]
|
| 48 |
|
| 49 |
-
# ===== ML FUNCTIONS =====
|
| 50 |
def save_cropped_images(original_image, boxes, labels, scores):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 51 |
saved_paths = []
|
|
|
|
| 52 |
for i, (box, label, score) in enumerate(zip(boxes, labels, scores)):
|
|
|
|
| 53 |
with tempfile.NamedTemporaryFile(suffix=".jpg", delete=False) as tmp_file:
|
| 54 |
filepath = tmp_file.name
|
|
|
|
|
|
|
| 55 |
cropped_img = original_image.crop(box)
|
|
|
|
|
|
|
| 56 |
cropped_img.save(filepath)
|
| 57 |
saved_paths.append(filepath)
|
|
|
|
| 58 |
return saved_paths
|
| 59 |
|
| 60 |
-
def draw_boxes(image, boxes, labels, scores, colors=BOX_COLORS, font_size=16):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 61 |
colour_cycle = cycle(colors)
|
| 62 |
draw = ImageDraw.Draw(image)
|
| 63 |
-
|
|
|
|
| 64 |
try:
|
| 65 |
-
font = ImageFont.truetype(
|
| 66 |
-
except:
|
| 67 |
-
font = ImageFont.load_default()
|
| 68 |
-
|
|
|
|
| 69 |
label_to_colour = {}
|
| 70 |
-
|
| 71 |
for box, label, score in zip(boxes, labels, scores):
|
|
|
|
| 72 |
colour = label_to_colour.setdefault(label, next(colour_cycle))
|
|
|
|
| 73 |
x_min, y_min, x_max, y_max = map(int, box)
|
| 74 |
-
|
|
|
|
| 75 |
draw.rectangle([x_min, y_min, x_max, y_max], outline=colour, width=2)
|
|
|
|
|
|
|
| 76 |
text = f"{label} ({score:.3f})"
|
| 77 |
-
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
|
| 81 |
-
|
| 82 |
draw.rectangle(bg_coords, fill=colour)
|
| 83 |
-
|
| 84 |
-
|
|
|
|
|
|
|
|
|
|
| 85 |
return image
|
| 86 |
|
| 87 |
-
def resize_image_max_dimension(image, max_size=
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 88 |
width, height = image.size
|
|
|
|
|
|
|
| 89 |
if max(width, height) <= max_size:
|
| 90 |
return image
|
|
|
|
|
|
|
| 91 |
ratio = max_size / max(width, height)
|
| 92 |
new_width = int(width * ratio)
|
| 93 |
new_height = int(height * ratio)
|
|
|
|
|
|
|
| 94 |
return image.resize((new_width, new_height), Image.Resampling.LANCZOS)
|
| 95 |
|
| 96 |
-
def detect_and_draw(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 97 |
text_query = text_query.lower()
|
| 98 |
-
|
| 99 |
-
|
|
|
|
|
|
|
|
|
|
| 100 |
inputs = processor(images=img, text=text_query, return_tensors="pt").to(DEVICE)
|
| 101 |
-
|
| 102 |
with torch.no_grad():
|
| 103 |
outputs = model(**inputs)
|
| 104 |
-
|
| 105 |
results = processor.post_process_grounded_object_detection(
|
| 106 |
outputs,
|
| 107 |
inputs.input_ids,
|
|
|
|
| 108 |
text_threshold=text_threshold,
|
| 109 |
target_sizes=[img.size[::-1]]
|
| 110 |
)[0]
|
| 111 |
-
|
| 112 |
img_out = img.copy()
|
| 113 |
img_out = draw_boxes(
|
| 114 |
img_out,
|
| 115 |
-
boxes=results["boxes"].cpu().numpy(),
|
| 116 |
-
labels=results.get("text_labels", results.get("labels", [])),
|
| 117 |
-
scores=results["scores"]
|
| 118 |
)
|
| 119 |
|
| 120 |
-
|
| 121 |
-
|
| 122 |
-
boxes=results["boxes"].cpu().numpy(),
|
| 123 |
-
labels=results.get("text_labels", results.get("labels", [])),
|
| 124 |
-
scores=results["scores"]
|
| 125 |
-
)
|
| 126 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 127 |
return img_out, crop_paths
|
| 128 |
|
| 129 |
-
#
|
| 130 |
-
|
| 131 |
-
|
| 132 |
-
|
| 133 |
-
return render_template_string('''
|
| 134 |
-
<!DOCTYPE html>
|
| 135 |
-
<html>
|
| 136 |
-
<head>
|
| 137 |
-
<title>Student Finder - Protetto</title>
|
| 138 |
-
<style>
|
| 139 |
-
body { font-family: Arial, sans-serif; max-width: 1200px; margin: 0 auto; padding: 20px; }
|
| 140 |
-
.header { background: #e8f5e8; padding: 20px; border-radius: 10px; margin-bottom: 20px; }
|
| 141 |
-
.content { background: #f5f5f5; padding: 30px; border-radius: 10px; }
|
| 142 |
-
.form-group { margin-bottom: 15px; }
|
| 143 |
-
label { display: block; margin-bottom: 5px; font-weight: bold; }
|
| 144 |
-
input, textarea, select { width: 100%; padding: 8px; border: 1px solid #ddd; border-radius: 4px; }
|
| 145 |
-
button { background: #007bff; color: white; padding: 10px 20px; border: none; border-radius: 4px; cursor: pointer; }
|
| 146 |
-
button:hover { background: #0056b3; }
|
| 147 |
-
.logout { float: right; }
|
| 148 |
-
.results { margin-top: 20px; }
|
| 149 |
-
.gallery { display: grid; grid-template-columns: repeat(auto-fit, minmax(200px, 1fr)); gap: 10px; margin-top: 20px; }
|
| 150 |
-
.gallery img { max-width: 100%; height: auto; border: 1px solid #ddd; border-radius: 4px; }
|
| 151 |
-
</style>
|
| 152 |
-
</head>
|
| 153 |
-
<body>
|
| 154 |
-
<div class="header">
|
| 155 |
-
<h1>🎓 Student Finder</h1>
|
| 156 |
-
<p>Carica una foto di classe e trova gli studenti</p>
|
| 157 |
-
<a href="/logout" class="logout">🔓 Logout</a>
|
| 158 |
-
<div style="clear: both;"></div>
|
| 159 |
-
</div>
|
| 160 |
-
|
| 161 |
-
<div class="content">
|
| 162 |
-
<form method="post" enctype="multipart/form-data" action="/detect">
|
| 163 |
-
<div class="form-group">
|
| 164 |
-
<label for="image">Immagine:</label>
|
| 165 |
-
<input type="file" id="image" name="image" accept="image/*" required>
|
| 166 |
-
</div>
|
| 167 |
-
|
| 168 |
-
<div class="form-group">
|
| 169 |
-
<label for="text_query">Text Query:</label>
|
| 170 |
-
<textarea id="text_query" name="text_query" rows="2" required>heads.</textarea>
|
| 171 |
-
<small>Testo in lowercase, ogni concetto termina con '.' (es. 'heads. faces.')</small>
|
| 172 |
-
</div>
|
| 173 |
-
|
| 174 |
-
<div class="form-group">
|
| 175 |
-
<label for="box_threshold">Box Threshold ({{ box_threshold }}):</label>
|
| 176 |
-
<input type="range" id="box_threshold" name="box_threshold" min="0" max="1" step="0.05" value="0.14">
|
| 177 |
-
</div>
|
| 178 |
-
|
| 179 |
-
<div class="form-group">
|
| 180 |
-
<label for="text_threshold">Text Threshold ({{ text_threshold }}):</label>
|
| 181 |
-
<input type="range" id="text_threshold" name="text_threshold" min="0" max="1" step="0.05" value="0.13">
|
| 182 |
-
</div>
|
| 183 |
-
|
| 184 |
-
<button type="submit">🔍 Rileva Studenti</button>
|
| 185 |
-
</form>
|
| 186 |
-
|
| 187 |
-
{% if result_image %}
|
| 188 |
-
<div class="results">
|
| 189 |
-
<h3>Risultati:</h3>
|
| 190 |
-
<img src="data:image/jpeg;base64,{{ result_image }}" alt="Risultato" style="max-width: 100%;">
|
| 191 |
-
|
| 192 |
-
{% if crops %}
|
| 193 |
-
<h4>Ritagli individuati ({{ crops|length }}):</h4>
|
| 194 |
-
<div class="gallery">
|
| 195 |
-
{% for crop in crops %}
|
| 196 |
-
<img src="data:image/jpeg;base64,{{ crop }}" alt="Ritaglio {{ loop.index }}">
|
| 197 |
-
{% endfor %}
|
| 198 |
-
</div>
|
| 199 |
-
{% endif %}
|
| 200 |
-
</div>
|
| 201 |
-
{% endif %}
|
| 202 |
-
</div>
|
| 203 |
-
</body>
|
| 204 |
-
</html>
|
| 205 |
-
''', box_threshold=0.14, text_threshold=0.13)
|
| 206 |
-
|
| 207 |
-
@app.route('/detect', methods=['POST'])
|
| 208 |
-
@require_auth
|
| 209 |
-
def detect():
|
| 210 |
-
if 'image' not in request.files:
|
| 211 |
-
return redirect(url_for('index'))
|
| 212 |
|
| 213 |
-
|
| 214 |
-
|
| 215 |
-
|
|
|
|
| 216 |
|
| 217 |
-
|
| 218 |
-
|
| 219 |
-
|
| 220 |
-
|
| 221 |
-
|
| 222 |
-
|
| 223 |
-
|
| 224 |
-
|
| 225 |
-
|
| 226 |
-
|
| 227 |
-
|
| 228 |
-
|
| 229 |
-
|
| 230 |
-
|
| 231 |
-
|
| 232 |
-
|
| 233 |
-
|
| 234 |
-
|
| 235 |
-
|
| 236 |
-
|
| 237 |
-
|
| 238 |
-
|
| 239 |
-
|
| 240 |
-
|
| 241 |
-
|
| 242 |
-
|
| 243 |
-
|
| 244 |
-
|
| 245 |
-
|
| 246 |
-
|
| 247 |
-
<head>
|
| 248 |
-
<title>Risultati - Student Finder</title>
|
| 249 |
-
<style>
|
| 250 |
-
body { font-family: Arial, sans-serif; max-width: 1200px; margin: 0 auto; padding: 20px; }
|
| 251 |
-
.header { background: #e8f5e8; padding: 20px; border-radius: 10px; margin-bottom: 20px; }
|
| 252 |
-
.content { background: #f5f5f5; padding: 30px; border-radius: 10px; }
|
| 253 |
-
.logout { float: right; }
|
| 254 |
-
.gallery { display: grid; grid-template-columns: repeat(auto-fit, minmax(200px, 1fr)); gap: 10px; margin-top: 20px; }
|
| 255 |
-
.gallery img { max-width: 100%; height: auto; border: 1px solid #ddd; border-radius: 4px; }
|
| 256 |
-
.back-btn { background: #6c757d; color: white; padding: 10px 20px; border: none; border-radius: 4px; cursor: pointer; text-decoration: none; display: inline-block; margin-bottom: 20px; }
|
| 257 |
-
.back-btn:hover { background: #545b62; }
|
| 258 |
-
</style>
|
| 259 |
-
</head>
|
| 260 |
-
<body>
|
| 261 |
-
<div class="header">
|
| 262 |
-
<h1>🎓 Risultati Student Finder</h1>
|
| 263 |
-
<a href="/logout" class="logout">🔓 Logout</a>
|
| 264 |
-
<div style="clear: both;"></div>
|
| 265 |
-
</div>
|
| 266 |
-
|
| 267 |
-
<a href="/" class="back-btn">← Nuova Analisi</a>
|
| 268 |
-
|
| 269 |
-
<div class="content">
|
| 270 |
-
<h3>Immagine con bounding box:</h3>
|
| 271 |
-
<img src="data:image/jpeg;base64,{{ result_image }}" alt="Risultato" style="max-width: 100%; border: 1px solid #ddd; border-radius: 4px;">
|
| 272 |
-
|
| 273 |
-
{% if crops %}
|
| 274 |
-
<h3>Ritagli individuati ({{ crops|length }}):</h3>
|
| 275 |
-
<div class="gallery">
|
| 276 |
-
{% for crop in crops %}
|
| 277 |
-
<img src="data:image/jpeg;base64,{{ crop }}" alt="Ritaglio {{ loop.index }}">
|
| 278 |
-
{% endfor %}
|
| 279 |
-
</div>
|
| 280 |
-
{% else %}
|
| 281 |
-
<p>Nessun ritaglio individuato.</p>
|
| 282 |
-
{% endif %}
|
| 283 |
-
</div>
|
| 284 |
-
</body>
|
| 285 |
-
</html>
|
| 286 |
-
''', result_image=result_b64, crops=crops_b64)
|
| 287 |
-
|
| 288 |
-
except Exception as e:
|
| 289 |
-
return f"Errore durante l'elaborazione: {str(e)}", 500
|
| 290 |
|
| 291 |
-
|
| 292 |
-
def
|
| 293 |
-
|
| 294 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 295 |
|
| 296 |
-
|
| 297 |
-
|
| 298 |
-
|
| 299 |
-
|
| 300 |
-
|
| 301 |
-
|
| 302 |
-
|
| 303 |
-
|
| 304 |
|
| 305 |
-
|
| 306 |
-
|
| 307 |
-
|
| 308 |
-
|
| 309 |
-
|
| 310 |
-
|
| 311 |
-
body {
|
| 312 |
-
font-family: Arial, sans-serif;
|
| 313 |
-
max-width: 400px;
|
| 314 |
-
margin: 100px auto;
|
| 315 |
-
padding: 20px;
|
| 316 |
-
background: #f5f5f5;
|
| 317 |
-
}
|
| 318 |
-
.login-form {
|
| 319 |
-
background: white;
|
| 320 |
-
padding: 30px;
|
| 321 |
-
border-radius: 10px;
|
| 322 |
-
box-shadow: 0 2px 10px rgba(0,0,0,0.1);
|
| 323 |
-
}
|
| 324 |
-
h2 {
|
| 325 |
-
color: #333;
|
| 326 |
-
text-align: center;
|
| 327 |
-
margin-bottom: 20px;
|
| 328 |
-
}
|
| 329 |
-
input[type="password"] {
|
| 330 |
-
width: 100%;
|
| 331 |
-
padding: 12px;
|
| 332 |
-
margin: 15px 0;
|
| 333 |
-
border: 1px solid #ddd;
|
| 334 |
-
border-radius: 5px;
|
| 335 |
-
box-sizing: border-box;
|
| 336 |
-
font-size: 16px;
|
| 337 |
-
}
|
| 338 |
-
button {
|
| 339 |
-
background: #007bff;
|
| 340 |
-
color: white;
|
| 341 |
-
padding: 12px 20px;
|
| 342 |
-
border: none;
|
| 343 |
-
border-radius: 5px;
|
| 344 |
-
cursor: pointer;
|
| 345 |
-
width: 100%;
|
| 346 |
-
font-size: 16px;
|
| 347 |
-
}
|
| 348 |
-
button:hover {
|
| 349 |
-
background: #0056b3;
|
| 350 |
-
}
|
| 351 |
-
.error {
|
| 352 |
-
color: red;
|
| 353 |
-
margin-bottom: 15px;
|
| 354 |
-
text-align: center;
|
| 355 |
-
padding: 10px;
|
| 356 |
-
background: #ffe6e6;
|
| 357 |
-
border-radius: 5px;
|
| 358 |
-
}
|
| 359 |
-
</style>
|
| 360 |
-
</head>
|
| 361 |
-
<body>
|
| 362 |
-
<div class="login-form">
|
| 363 |
-
<h2>🔒 Student Finder - Accesso Protetto</h2>
|
| 364 |
-
<p style="text-align: center; color: #666;">Inserisci la password per accedere</p>
|
| 365 |
-
{% if error %}
|
| 366 |
-
<div class="error">{{ error }}</div>
|
| 367 |
-
{% endif %}
|
| 368 |
-
<form method="POST">
|
| 369 |
-
<input type="password" name="password" placeholder="Password" required>
|
| 370 |
-
<button type="submit">🔑 Accedi</button>
|
| 371 |
-
</form>
|
| 372 |
-
</div>
|
| 373 |
-
</body>
|
| 374 |
-
</html>
|
| 375 |
-
''', error=error)
|
| 376 |
-
|
| 377 |
-
@app.route('/logout')
|
| 378 |
-
def logout():
|
| 379 |
-
session.clear()
|
| 380 |
-
return redirect(url_for('login'))
|
| 381 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 382 |
|
| 383 |
-
if __name__ ==
|
| 384 |
-
|
| 385 |
-
app.run(host='0.0.0.0', port=port, debug=False)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
import torch
|
| 2 |
+
from PIL import Image, ImageDraw, ImageFont
|
| 3 |
from transformers import GroundingDinoProcessor
|
| 4 |
from modeling_grounding_dino import GroundingDinoForObjectDetection
|
| 5 |
+
|
| 6 |
+
from PIL import Image, ImageDraw, ImageFont
|
| 7 |
from itertools import cycle
|
| 8 |
+
import os
|
| 9 |
+
from datetime import datetime
|
| 10 |
+
import gradio as gr
|
| 11 |
import tempfile
|
| 12 |
+
|
| 13 |
+
# Load model and processor
|
| 14 |
+
model_id = "fushh7/llmdet_swin_large_hf"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 15 |
model_id = "fushh7/llmdet_swin_tiny_hf"
|
| 16 |
+
DEVICE = "cpu"
|
| 17 |
|
| 18 |
print(f"[INFO] Using device: {DEVICE}")
|
| 19 |
print(f"[INFO] Loading model from {model_id}...")
|
|
|
|
| 24 |
|
| 25 |
print("[INFO] Model loaded successfully.")
|
| 26 |
|
| 27 |
+
# Pre-defined palette (extend or tweak as you like)
|
| 28 |
BOX_COLORS = [
|
| 29 |
"deepskyblue", "red", "lime", "dodgerblue",
|
| 30 |
+
"cyan", "magenta", "yellow",
|
| 31 |
+
"orange", "chartreuse"
|
| 32 |
]
|
| 33 |
|
|
|
|
| 34 |
def save_cropped_images(original_image, boxes, labels, scores):
|
| 35 |
+
"""
|
| 36 |
+
Salva ogni regione ritagliata definita dalle bounding box in file temporanei.
|
| 37 |
+
|
| 38 |
+
:param original_image: Immagine PIL originale
|
| 39 |
+
:param boxes: Lista di bounding box [x_min, y_min, x_max, y_max]
|
| 40 |
+
:param labels: Lista di etichette per ogni box
|
| 41 |
+
:param scores: Lista di punteggi di confidenza
|
| 42 |
+
:return: Lista dei percorsi dei file temporanei salvati
|
| 43 |
+
"""
|
| 44 |
saved_paths = []
|
| 45 |
+
|
| 46 |
for i, (box, label, score) in enumerate(zip(boxes, labels, scores)):
|
| 47 |
+
# Crea un file temporaneo
|
| 48 |
with tempfile.NamedTemporaryFile(suffix=".jpg", delete=False) as tmp_file:
|
| 49 |
filepath = tmp_file.name
|
| 50 |
+
|
| 51 |
+
# Ritaglia la regione dall'immagine originale
|
| 52 |
cropped_img = original_image.crop(box)
|
| 53 |
+
|
| 54 |
+
# Salva l'immagine ritagliata
|
| 55 |
cropped_img.save(filepath)
|
| 56 |
saved_paths.append(filepath)
|
| 57 |
+
|
| 58 |
return saved_paths
|
| 59 |
|
| 60 |
+
def draw_boxes(image, boxes, labels, scores, colors=BOX_COLORS, font_path="arial.ttf", font_size=16):
|
| 61 |
+
"""
|
| 62 |
+
Draw bounding boxes and labels on a PIL Image.
|
| 63 |
+
|
| 64 |
+
:param image: PIL Image object
|
| 65 |
+
:param boxes: Iterable of [x_min, y_min, x_max, y_max]
|
| 66 |
+
:param labels: Iterable of label strings
|
| 67 |
+
:param scores: Iterable of scalar confidences (0-1)
|
| 68 |
+
:param colors: List/tuple of colour names or RGB tuples
|
| 69 |
+
:param font_path: Path to a TTF font for labels
|
| 70 |
+
:param font_size: Int size of font to use, default 16
|
| 71 |
+
:return: PIL Image with drawn boxes
|
| 72 |
+
"""
|
| 73 |
+
# Ensure we can iterate colours indefinitely
|
| 74 |
colour_cycle = cycle(colors)
|
| 75 |
draw = ImageDraw.Draw(image)
|
| 76 |
+
|
| 77 |
+
# Pick a font (fallback to default if missing)
|
| 78 |
try:
|
| 79 |
+
font = ImageFont.truetype(font_path, size=font_size)
|
| 80 |
+
except IOError:
|
| 81 |
+
font = ImageFont.load_default(size=font_size)
|
| 82 |
+
|
| 83 |
+
# Assign a consistent colour per label (optional)
|
| 84 |
label_to_colour = {}
|
| 85 |
+
|
| 86 |
for box, label, score in zip(boxes, labels, scores):
|
| 87 |
+
# Reuse colour if label seen before, else take next from cycle
|
| 88 |
colour = label_to_colour.setdefault(label, next(colour_cycle))
|
| 89 |
+
|
| 90 |
x_min, y_min, x_max, y_max = map(int, box)
|
| 91 |
+
|
| 92 |
+
# Draw rectangle
|
| 93 |
draw.rectangle([x_min, y_min, x_max, y_max], outline=colour, width=2)
|
| 94 |
+
|
| 95 |
+
# Compose text
|
| 96 |
text = f"{label} ({score:.3f})"
|
| 97 |
+
text_size = draw.textbbox((0, 0), text, font=font)[2:]
|
| 98 |
+
|
| 99 |
+
# Draw text background for legibility
|
| 100 |
+
bg_coords = [x_min, y_min - text_size[1] - 4,
|
| 101 |
+
x_min + text_size[0] + 4, y_min]
|
| 102 |
draw.rectangle(bg_coords, fill=colour)
|
| 103 |
+
|
| 104 |
+
# Draw text
|
| 105 |
+
draw.text((x_min + 2, y_min - text_size[1] - 2),
|
| 106 |
+
text, fill="black", font=font)
|
| 107 |
+
|
| 108 |
return image
|
| 109 |
|
| 110 |
+
def resize_image_max_dimension(image, max_size=4096):
|
| 111 |
+
"""
|
| 112 |
+
Resize an image so that the longest side is at most max_size pixels,
|
| 113 |
+
while maintaining the aspect ratio.
|
| 114 |
+
|
| 115 |
+
:param image: PIL Image object
|
| 116 |
+
:param max_size: Maximum dimension in pixels (default: 1024)
|
| 117 |
+
:return: PIL Image object (resized)
|
| 118 |
+
"""
|
| 119 |
width, height = image.size
|
| 120 |
+
|
| 121 |
+
# Check if resizing is needed
|
| 122 |
if max(width, height) <= max_size:
|
| 123 |
return image
|
| 124 |
+
|
| 125 |
+
# Calculate new dimensions maintaining aspect ratio
|
| 126 |
ratio = max_size / max(width, height)
|
| 127 |
new_width = int(width * ratio)
|
| 128 |
new_height = int(height * ratio)
|
| 129 |
+
|
| 130 |
+
# Resize the image using high-quality resampling
|
| 131 |
return image.resize((new_width, new_height), Image.Resampling.LANCZOS)
|
| 132 |
|
| 133 |
+
def detect_and_draw(
|
| 134 |
+
img: Image.Image,
|
| 135 |
+
text_query: str,
|
| 136 |
+
box_threshold: float = 0.14,
|
| 137 |
+
text_threshold: float = 0.13,
|
| 138 |
+
save_crops: bool = True
|
| 139 |
+
):
|
| 140 |
+
"""
|
| 141 |
+
Detect objects described in `text_query`, draw boxes, return the image and crops.
|
| 142 |
+
Note: `text_query` must be lowercase and each concept ends with a dot
|
| 143 |
+
(e.g. 'a cat. a remote control.')
|
| 144 |
+
"""
|
| 145 |
+
|
| 146 |
+
# Make sure text is lowered
|
| 147 |
text_query = text_query.lower()
|
| 148 |
+
|
| 149 |
+
# If the image size is too large, we make it smaller
|
| 150 |
+
img = resize_image_max_dimension(img, max_size=4096)
|
| 151 |
+
|
| 152 |
+
# Preprocess the image
|
| 153 |
inputs = processor(images=img, text=text_query, return_tensors="pt").to(DEVICE)
|
| 154 |
+
|
| 155 |
with torch.no_grad():
|
| 156 |
outputs = model(**inputs)
|
| 157 |
+
|
| 158 |
results = processor.post_process_grounded_object_detection(
|
| 159 |
outputs,
|
| 160 |
inputs.input_ids,
|
| 161 |
+
box_threshold=box_threshold,
|
| 162 |
text_threshold=text_threshold,
|
| 163 |
target_sizes=[img.size[::-1]]
|
| 164 |
)[0]
|
| 165 |
+
|
| 166 |
img_out = img.copy()
|
| 167 |
img_out = draw_boxes(
|
| 168 |
img_out,
|
| 169 |
+
boxes = results["boxes"].cpu().numpy(),
|
| 170 |
+
labels = results.get("text_labels", results.get("labels", [])),
|
| 171 |
+
scores = results["scores"]
|
| 172 |
)
|
| 173 |
|
| 174 |
+
# Lista per i percorsi dei crop
|
| 175 |
+
crop_paths = []
|
|
|
|
|
|
|
|
|
|
|
|
|
| 176 |
|
| 177 |
+
if save_crops:
|
| 178 |
+
crop_paths = save_cropped_images(
|
| 179 |
+
img,
|
| 180 |
+
boxes=results["boxes"].cpu().numpy(),
|
| 181 |
+
labels=results.get("text_labels", results.get("labels", [])),
|
| 182 |
+
scores=results["scores"]
|
| 183 |
+
)
|
| 184 |
+
print(f"Generated {len(crop_paths)} cropped images")
|
| 185 |
+
|
| 186 |
return img_out, crop_paths
|
| 187 |
|
| 188 |
+
# Create example list dynamically from examples directory
|
| 189 |
+
def load_examples_from_directory(directory="examples"):
|
| 190 |
+
"""
|
| 191 |
+
Carica automaticamente tutti i file JPG dalla directory degli esempi.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 192 |
|
| 193 |
+
:param directory: Percorso della directory contenente gli esempi
|
| 194 |
+
:return: Lista di esempi nel formato [filepath, text_query, box_threshold, text_threshold]
|
| 195 |
+
"""
|
| 196 |
+
examples = []
|
| 197 |
|
| 198 |
+
# Verifica se la directory esiste
|
| 199 |
+
if not os.path.exists(directory):
|
| 200 |
+
print(f"[WARNING] Directory '{directory}' non trovata. Creala e aggiungi file JPG.")
|
| 201 |
+
return examples
|
| 202 |
+
|
| 203 |
+
# Cerca tutti i file JPG nella directory
|
| 204 |
+
#jpg_files = [f for f in os.listdir(directory) if f.lower().endswith('.jpg')]
|
| 205 |
+
jpg_files = [f for f in os.listdir(directory) if f.lower().endswith(('.jpg', '.png'))]
|
| 206 |
+
if not jpg_files:
|
| 207 |
+
print(f"[WARNING] Nessun file JPG trovato nella directory '{directory}'")
|
| 208 |
+
return examples
|
| 209 |
+
|
| 210 |
+
print(f"[INFO] Trovati {len(jpg_files)} file JPG nella directory examples/")
|
| 211 |
+
|
| 212 |
+
# Crea gli esempi per ogni file JPG
|
| 213 |
+
for jpg_file in jpg_files:
|
| 214 |
+
filepath = os.path.join(directory, jpg_file)
|
| 215 |
+
examples.append([filepath, "heads.", 0.24, 0.23])
|
| 216 |
+
|
| 217 |
+
return examples
|
| 218 |
+
|
| 219 |
+
# Popola automaticamente la lista degli esempi
|
| 220 |
+
examples = load_examples_from_directory()
|
| 221 |
+
|
| 222 |
+
# Se non sono stati trovati esempi, usa un esempio di fallback
|
| 223 |
+
if not examples:
|
| 224 |
+
print("[INFO] Usando esempio di fallback")
|
| 225 |
+
examples = [
|
| 226 |
+
["examples/stickers(1).jpg", "heads.", 0.24, 0.23],
|
| 227 |
+
]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 228 |
|
| 229 |
+
# Funzione per pulire i file temporanei dopo l'uso
|
| 230 |
+
def cleanup_temp_files(crop_paths):
|
| 231 |
+
for path in crop_paths:
|
| 232 |
+
try:
|
| 233 |
+
os.unlink(path)
|
| 234 |
+
except:
|
| 235 |
+
pass
|
| 236 |
+
|
| 237 |
+
# Create Gradio demo
|
| 238 |
+
with gr.Blocks(title="ClasmateFaceFinder", css=".gradio-container {max-width: 100% !important}") as demo:
|
| 239 |
+
gr.Markdown("# Classmate Finder")
|
| 240 |
+
gr.Markdown("Upload an image and adjust thresholds to see detections.")
|
| 241 |
+
|
| 242 |
+
with gr.Row():
|
| 243 |
+
with gr.Column():
|
| 244 |
+
image_input = gr.Image(type="pil", label="Input Image")
|
| 245 |
+
text_query = gr.Textbox(
|
| 246 |
+
value="head.",
|
| 247 |
+
label="Text Query (lowercase, end each with '.', for example 'a bird. a tree.')"
|
| 248 |
+
)
|
| 249 |
+
box_threshold = gr.Slider(0.0, 1.0, 0.14, step=0.05, label="Box Threshold")
|
| 250 |
+
text_threshold = gr.Slider(0.0, 1.0, 0.13, step=0.05, label="Text Threshold")
|
| 251 |
+
submit_btn = gr.Button("Detect")
|
| 252 |
+
|
| 253 |
+
with gr.Column():
|
| 254 |
+
image_output = gr.Image(type="pil", label="Detections")
|
| 255 |
+
|
| 256 |
+
# Galleria per i crop
|
| 257 |
+
gallery = gr.Gallery(
|
| 258 |
+
label="Detected Crops",
|
| 259 |
+
columns=[4],
|
| 260 |
+
rows=[2],
|
| 261 |
+
object_fit="contain",
|
| 262 |
+
height="auto"
|
| 263 |
+
)
|
| 264 |
|
| 265 |
+
# Esempi
|
| 266 |
+
gr.Examples(
|
| 267 |
+
examples=examples,
|
| 268 |
+
inputs=[image_input, text_query, box_threshold, text_threshold],
|
| 269 |
+
outputs=[image_output, gallery],
|
| 270 |
+
fn=detect_and_draw,
|
| 271 |
+
cache_examples=True
|
| 272 |
+
)
|
| 273 |
|
| 274 |
+
# Pulsante di submit
|
| 275 |
+
submit_btn.click(
|
| 276 |
+
fn=detect_and_draw,
|
| 277 |
+
inputs=[image_input, text_query, box_threshold, text_threshold],
|
| 278 |
+
outputs=[image_output, gallery]
|
| 279 |
+
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 280 |
|
| 281 |
+
# Pulisci i file temporanei quando viene caricato un nuovo esempio
|
| 282 |
+
demo.load(
|
| 283 |
+
fn=lambda: None,
|
| 284 |
+
inputs=None,
|
| 285 |
+
outputs=None,
|
| 286 |
+
)
|
| 287 |
|
| 288 |
+
if __name__ == "__main__":
|
| 289 |
+
demo.launch(server_name="0.0.0.0", share=False)
|
|
|