Spaces:
Sleeping
Sleeping
File size: 19,660 Bytes
bd5df39 1f6e06f d130a53 bd5df39 9010dc6 d130a53 bd5df39 d130a53 bd5df39 9010dc6 c275fbc 9010dc6 c275fbc 021aa52 c275fbc 021aa52 c275fbc 021aa52 c275fbc 021aa52 c275fbc 021aa52 c275fbc 021aa52 c275fbc 021aa52 c275fbc 021aa52 c275fbc 021aa52 c275fbc 021aa52 9010dc6 c275fbc 021aa52 c275fbc 9010dc6 27e651e d130a53 bd5df39 27e651e 9010dc6 021aa52 9010dc6 27e651e d130a53 bd5df39 d130a53 bd5df39 d130a53 bd5df39 27e651e d130a53 bd5df39 d130a53 27e651e d130a53 27e651e d130a53 9010dc6 021aa52 d130a53 021aa52 27e651e 9010dc6 d130a53 bd5df39 1f6e06f 27e651e d130a53 bd5df39 d130a53 bd5df39 d130a53 27e651e d130a53 bd5df39 d130a53 bd5df39 d130a53 bd5df39 27e651e d130a53 bd5df39 27e651e bd5df39 27e651e d130a53 bd5df39 d130a53 bd5df39 d130a53 bd5df39 d130a53 bd5df39 d130a53 bd5df39 d130a53 021aa52 d130a53 a8169fb d130a53 1f6e06f d130a53 c275fbc d130a53 c275fbc d130a53 c275fbc 9010dc6 d130a53 9010dc6 a5bb7e3 9010dc6 021aa52 9010dc6 d130a53 021aa52 c275fbc d130a53 a5bb7e3 d130a53 c275fbc 27e651e d130a53 27e651e d130a53 bd5df39 c275fbc d130a53 c275fbc bd5df39 d130a53 c275fbc 27e651e d130a53 bd5df39 9010dc6 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 | import os
import uuid
import cv2
import torch
import torch.nn as nn
import numpy as np
import datetime
import threading
import httpx
import time
import asyncio
from collections import deque
from fastapi import FastAPI, UploadFile, File, BackgroundTasks, HTTPException, Query, Request
from fastapi.responses import FileResponse, RedirectResponse, StreamingResponse
from pydantic import BaseModel
from pytorchvideo.models.hub import slowfast_r50
from ultralytics import YOLO
from dotenv import load_dotenv
load_dotenv(override=True)
# --- DOCUMENTATION METADATA ---
tags_metadata = [
{
"name": "Live Stream Monitoring",
"description": "Real-time AI surveillance for RTSP cameras.",
},
{
"name": "Recorded Video Analysis",
"description": "Upload and process video files for theft detection.",
},
]
description = """
<details>
<summary><b> USER GUIDE & ENDPOINT MANUAL (Click to Expand)</b></summary>
<br>
### Section 1: Live Stream Monitoring
*Use these endpoints to manage and watch live AI security feeds.*
1. **`POST /stream/create` (Register Camera)**
- **What it does:** Saves your camera's RTSP address into the system.
- **How to use:** Enter a name, the RTSP link, and the location.
2. **`POST /stream/start/{id}` (Activate AI)**
- **What it does:** Starts the AI processing.
- **How to use:** Returns stream URL to view the stream.
3. **`PUT /stream/update/{id}` (Change Settings)**
- **What it does:** Updates the name or RTSP link of an existing camera.
4. **`DELETE /stream/delete/{id}` (Remove Camera)**
- **What it does:** Stops the stream and removes the camera from the list.
5. **`GET /cameras/{id}/frame` (Main Video Feed)**
- **What it does:** The actual live video stream.
6. **`GET /stream/list_cameras` (Camera Directory)**
- **What it does:** Displays all added cameras and their status.
7. **`POST /stream/stop/{id}` (Deactivate AI)**
- **What it does:** Shuts down the AI for a camera.
<br>
### Section 2: Recorded Video Analysis
*Use these endpoints to scan uploaded files for theft incidents.*
1. **`POST /video/detect` (Upload for Analysis)**
- **What it does:** Uploads a video file for a full AI scan. Returns a **job_id**.
2. **`GET /video/status/{job_id}` (Check Progress)**
- **What it does:** Provides the scan percentage (0% to 100%).
3. **`GET /video/jobs` (Task History)**
- **What it does:** Lists every video ever uploaded.
4. **`GET /video/download/{job_id}` (Get Result)**
- **What it does:** Download the final processed video once status is 100%.
</details>
---
"""
# --- CONFIGURATION ---
DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
MODEL_PATH = "best_slowfast_theft.pth"
DISCORD_WEBHOOK_URL = os.getenv("DISCORD_WEBHOOK_URL")
UPLOAD_DIR = "uploads"
OUTPUT_DIR = "outputs"
CLIP_LEN = 32
IMG_SIZE = 224
THEFT_THRESHOLD = 0.6
os.makedirs(UPLOAD_DIR, exist_ok=True)
os.makedirs(OUTPUT_DIR, exist_ok=True)
app = FastAPI(title="AI Theft Detection Backend",
description=description,
openapi_tags=tags_metadata,
version="1.0.0")
# --- DATABASE & STATE ---
MOCK_DB = {"cameras": {}}
jobs = {}
# --- LOAD MODELS ---
print(f"Initializing Models on {DEVICE}...")
yolo = YOLO("yolov8n.pt")
slowfast_model = slowfast_r50(pretrained=False)
slowfast_model.blocks[-1].proj = nn.Sequential(
nn.Dropout(p=0.5),
nn.Linear(slowfast_model.blocks[-1].proj.in_features, 2)
)
if os.path.exists(MODEL_PATH):
ckpt = torch.load(MODEL_PATH, map_location=DEVICE)
slowfast_model.load_state_dict(ckpt["model"] if "model" in ckpt else ckpt)
slowfast_model = slowfast_model.to(DEVICE).eval()
# --- SCHEMAS ---
class CameraCreate(BaseModel):
name: str; rtspUrl: str; location: str
class CameraUpdate(BaseModel):
name: str = None; rtspUrl: str = None; location: str = None
# --- NOTIFICATION LOGIC ---
import json
async def send_discord_alert(source_name, score, crop_frame=None, mode="live"):
if not DISCORD_WEBHOOK_URL or "YOUR_DISCORD" in DISCORD_WEBHOOK_URL:
return
now = datetime.datetime.now().strftime("%Y-%m-%d %I:%M:%S %p")
filename = "theft_crop.jpg"
footer_text = "Security AI Alert System • Recorded video" if mode == "recorded" else "Security AI Alert System • Live Monitoring"
payload = {
"username": "Surveillance Monitoring",
"avatar_url": "https://cdn-icons-png.flaticon.com/512/2564/2564388.png",
"embeds": [{
"title": "🚨 THEFT DETECTED",
"description": f"AI System has flagged suspicious activity at **{source_name}**.",
"color": 15548997,
"fields": [
{"name": "📊 Confidence Score", "value": f"**{int(score*100)}%**", "inline": True},
{"name": "📍 Location", "value": f"**{source_name}**", "inline": True},
{"name": "⏱️ Timestamp", "value": f"**{now}**", "inline": False}
],
"image": {"url": f"attachment://{filename}"},
"footer": {"text": footer_text}
}]
}
try:
async with httpx.AsyncClient() as client:
if crop_frame is not None:
_, buffer = cv2.imencode('.jpg', crop_frame)
files = {
'file': (filename, buffer.tobytes(), 'image/jpeg')
}
await client.post(
DISCORD_WEBHOOK_URL,
data={"payload_json": json.dumps(payload)},
files=files
)
else:
await client.post(DISCORD_WEBHOOK_URL, json=payload)
except Exception as e:
print(f"Discord Alert Error: {e}")
# --- VISUALIZATION HELPERS ---
def draw_corner_rect(img, pt1, pt2, color, thickness=2, r=15, d=25):
x1, y1 = pt1
x2, y2 = pt2
cv2.line(img, (x1+r, y1), (x1+r+d, y1), color, thickness)
cv2.line(img, (x1, y1+r), (x1, y1+r+d), color, thickness)
cv2.ellipse(img, (x1+r, y1+r), (r,r), 180, 0, 90, color, thickness)
cv2.line(img, (x2-r, y1), (x2-r-d, y1), color, thickness)
cv2.line(img, (x2, y1+r), (x2, y1+r+d), color, thickness)
cv2.ellipse(img, (x2-r, y1+r), (r,r), 270, 0, 90, color, thickness)
cv2.line(img, (x1+r, y2), (x1+r+d, y2), color, thickness)
cv2.line(img, (x1, y2-r), (x1, y2-r-d), color, thickness)
cv2.ellipse(img, (x1+r, y2-r), (r,r), 90, 0, 90, color, thickness)
cv2.line(img, (x2-r, y2), (x2-r-d, y2), color, thickness)
cv2.line(img, (x2, y2-r), (x2, y2-r-d), color, thickness)
cv2.ellipse(img, (x2-r, y2-r), (r,r), 0, 0, 90, color, thickness)
def draw_security_card(frame, avg_prob, theft_flag, title="AI ANALYZER"):
card_x, card_y = 35, 35
padding = 30
line_spacing = 45
card_w, card_h = 620, 260
orange = (0, 165, 255)
bg_color = (25, 25, 25)
overlay = frame.copy()
cv2.rectangle(overlay, (card_x, card_y), (card_x + card_w, card_y + card_h), bg_color, -1)
cv2.addWeighted(overlay, 0.85, frame, 0.15, 0, frame)
cv2.rectangle(frame, (card_x, card_y), (card_x + card_w, card_y + card_h), orange, 3)
status_label = "ALERT: THEFT DETECTED" if theft_flag else "STATUS: SYSTEM NORMAL"
status_color = (0, 0, 255) if theft_flag else (0, 255, 0)
now = datetime.datetime.now().strftime("%b %d, %Y | %I:%M:%S %p")
confidence = f"AI CONFIDENCE: {int(avg_prob * 100)}%"
curr_y = card_y + padding + 15
header_text = "THEFT DETECTION LIVE MONITORING"
(tw, th), _ = cv2.getTextSize(header_text, cv2.FONT_HERSHEY_DUPLEX, 0.9, 2)
cv2.putText(frame, header_text, (card_x + (card_w - tw) // 2, curr_y),
cv2.FONT_HERSHEY_DUPLEX, 0.9, (255, 255, 255), 2, cv2.LINE_AA)
curr_y += line_spacing + 5
cv2.putText(frame, f"SOURCE: {title.upper()}", (card_x + padding, curr_y),
cv2.FONT_HERSHEY_DUPLEX, 0.8, (200, 200, 200), 1, cv2.LINE_AA)
curr_y += line_spacing
cv2.putText(frame, status_label, (card_x + padding, curr_y),
cv2.FONT_HERSHEY_DUPLEX, 1.0, status_color, 2, cv2.LINE_AA)
curr_y += line_spacing
cv2.putText(frame, confidence, (card_x + padding, curr_y),
cv2.FONT_HERSHEY_DUPLEX, 0.8, (255, 255, 255), 1, cv2.LINE_AA)
curr_y += line_spacing
cv2.putText(frame, now, (card_x + padding, curr_y),
cv2.FONT_HERSHEY_SIMPLEX, 0.6, (150, 150, 150), 1, cv2.LINE_AA)
# --- PROCESSING CORE ---
def preprocess(frames):
processed = [cv2.resize(f, (IMG_SIZE, IMG_SIZE))[:, :, ::-1] / 255.0 for f in frames]
clip = np.transpose(np.array(processed), (3, 0, 1, 2))
return torch.tensor(clip).float().unsqueeze(0).to(DEVICE)
# --- BACKGROUND TASK: VIDEO FILE ---
async def process_video_file(job_id, in_p, out_p):
try:
cap = cv2.VideoCapture(in_p)
w, h = int(cap.get(3)), int(cap.get(4))
fps, total_frames = int(cap.get(5)), int(cap.get(7))
out = cv2.VideoWriter(out_p, cv2.VideoWriter_fourcc(*"mp4v"), fps, (w, h))
frame_buffer = deque(maxlen=CLIP_LEN)
prediction_buffer = deque(maxlen=10)
last_alert_time = 0
frame_counter = 0
source_name = jobs.get(job_id, {}).get("filename") or f"Recorded Video ({job_id})"
while cap.isOpened():
if jobs.get(job_id, {}).get("stop_requested"): break
ret, frame = cap.read()
if not ret: break
frame_counter += 1
theft_flag, avg_prob, current_crop = False, 0.0, None
results = yolo(frame, verbose=False)
for r in results:
for box in r.boxes:
if int(box.cls[0]) == 0:
x1, y1, x2, y2 = map(int, box.xyxy[0])
current_crop = frame[y1:y2, x1:x2]
if current_crop.size > 0:
frame_buffer.append(current_crop)
if len(frame_buffer) == CLIP_LEN:
clip_ts = preprocess(frame_buffer)
with torch.no_grad():
probs = torch.softmax(slowfast_model([clip_ts[:,:,::4,:,:], clip_ts]), dim=1)
prediction_buffer.append(probs[0][1].item())
avg_prob = np.mean(prediction_buffer)
if avg_prob > THEFT_THRESHOLD:
theft_flag = True
draw_corner_rect(frame, (x1,y1), (x2,y2), (0,0,255))
current_time = time.time()
if current_time - last_alert_time > 30 and current_crop is not None:
await send_discord_alert(source_name, avg_prob, current_crop, mode="recorded")
last_alert_time = current_time
else:
draw_corner_rect(frame, (x1,y1), (x2,y2), (0,255,0))
draw_security_card(frame, avg_prob, theft_flag, "VIDEO ANALYSIS")
out.write(frame)
jobs[job_id]["progress"] = int((frame_counter / total_frames) * 100)
cap.release(); out.release()
jobs[job_id]["status"] = "completed" if not jobs[job_id].get("stop_requested") else "stopped"
except Exception as e:
jobs[job_id]["status"] = f"failed: {str(e)}"
# --- RTSP CAMERA PIPELINE ---
class CameraPipeline:
def __init__(self, cam_id, name, url):
self.cam_id = cam_id
self.name = name
self.url = url
self.running = True
self.latest_frame = None
self.frame_buffer = deque(maxlen=CLIP_LEN)
self.prediction_buffer = deque(maxlen=10)
self.last_alert_time = 0
self.raw_frame = None
self.capture_thread = threading.Thread(target=self._capture_loop, daemon=True)
self.ai_thread = threading.Thread(target=self._ai_loop, daemon=True)
self.capture_thread.start()
self.ai_thread.start()
def _capture_loop(self):
os.environ["OPENCV_FFMPEG_CAPTURE_OPTIONS"] = "rtsp_transport;tcp"
cap = cv2.VideoCapture(self.url, cv2.CAP_FFMPEG)
cap.set(cv2.CAP_PROP_BUFFERSIZE, 1)
while self.running:
ret, frame = cap.read()
if not ret:
cap.release()
time.sleep(2)
cap = cv2.VideoCapture(self.url, cv2.CAP_FFMPEG)
continue
self.raw_frame = frame
cap.release()
def _ai_loop(self):
while self.running:
if self.raw_frame is None:
time.sleep(0.01)
continue
frame = self.raw_frame.copy()
theft_flag, avg_prob = False, 0.0
current_crop = None
results = yolo(frame, verbose=False, conf=0.4)
for r in results:
for box in r.boxes:
if int(box.cls[0]) == 0:
x1, y1, x2, y2 = map(int, box.xyxy[0])
current_crop = frame[y1:y2, x1:x2]
if current_crop.size > 0:
p_frame = cv2.resize(current_crop, (IMG_SIZE, IMG_SIZE))[:,:,::-1]/255.0
self.frame_buffer.append(p_frame)
if len(self.frame_buffer) == CLIP_LEN:
clip = np.transpose(np.array(self.frame_buffer), (3,0,1,2))
clip_ts = torch.tensor(clip).float().unsqueeze(0).to(DEVICE)
with torch.no_grad():
probs = torch.softmax(slowfast_model([clip_ts[:,:,::4,:,:], clip_ts]), dim=1)
self.prediction_buffer.append(probs[0][1].item())
avg_prob = np.mean(prediction_buffer)
is_theft = avg_prob > THEFT_THRESHOLD
draw_corner_rect(frame, (x1,y1), (x2,y2), (0,0,255) if is_theft else (0,255,0))
if is_theft:
theft_flag = True
if theft_flag and (time.time() - self.last_alert_time > 30) and current_crop is not None:
try:
asyncio.run(send_discord_alert(self.name, avg_prob, current_crop, mode="live"))
except RuntimeError:
loop = asyncio.new_event_loop()
loop.run_until_complete(send_discord_alert(self.name, avg_prob, current_crop, mode="live"))
loop.close()
self.last_alert_time = time.time()
draw_security_card(frame, avg_prob, theft_flag, self.name)
self.latest_frame = frame
class StreamManager:
def __init__(self):
self.active_pipelines = {}
def start_camera(self, cam_id, name, url):
if cam_id in self.active_pipelines:
return False
self.active_pipelines[cam_id] = CameraPipeline(cam_id, name, url)
return True
def stop_camera(self, cam_id):
if cam_id not in self.active_pipelines:
return False
self.active_pipelines[cam_id].running = False
del self.active_pipelines[cam_id]
return True
manager = StreamManager()
# --- ENDPOINTS ---
@app.get("/", include_in_schema=False)
async def root(): return RedirectResponse(url="/docs")
# --- CAMERA ENDPOINTS ---
@app.get("/stream/list_cameras", tags=["Live Stream Monitoring"])
def get_cameras(): return {"cameras": list(MOCK_DB["cameras"].values())}
@app.post("/stream/create", tags=["Live Stream Monitoring"])
def create_camera(cam: CameraCreate):
new_id = f"cam-{len(MOCK_DB['cameras']) + 1}"
camera = {**cam.dict(), "id": new_id, "status": "offline", "isStreaming": False}
MOCK_DB["cameras"][new_id] = camera
return {"camera": camera}
@app.post("/stream/start/{id}", tags=["Live Stream Monitoring"])
async def start_camera(id: str, request: Request):
if id not in MOCK_DB["cameras"]:
raise HTTPException(404)
cam = MOCK_DB["cameras"][id]
started = manager.start_camera(id, cam["name"], cam["rtspUrl"])
if started:
cam["status"] = "online"
cam["isStreaming"] = True
# Generate URLs for the user
local_url = f"{request.base_url}cameras/{id}/frame"
return {
"success": started,
"job_id": id,
"view_urls": {
"stream_url": str(local_url)
}
}
@app.put("/stream/update/{id}", tags=["Live Stream Monitoring"])
def update_camera(id: str, cam_data: CameraUpdate):
if id not in MOCK_DB["cameras"]:
raise HTTPException(status_code=404, detail="Camera not found")
current_cam = MOCK_DB["cameras"][id]
update_dict = cam_data.dict(exclude_unset=True)
current_cam.update(update_dict)
return {"message": "Camera updated successfully", "camera": current_cam}
@app.delete("/stream/delete/{id}", tags=["Live Stream Monitoring"])
def delete_camera(id: str):
if id not in MOCK_DB["cameras"]:
raise HTTPException(status_code=404, detail="Camera not found")
manager.stop_camera(id)
del MOCK_DB["cameras"][id]
return {"message": f"Camera {id} deleted successfully"}
@app.post("/stream/stop/{id}", tags=["Live Stream Monitoring"])
def stop_camera(id: str):
if id not in MOCK_DB["cameras"]:
raise HTTPException(404)
stopped = manager.stop_camera(id)
if stopped:
MOCK_DB["cameras"][id]["status"] = "offline"
MOCK_DB["cameras"][id]["isStreaming"] = False
return {"success": stopped}
@app.get("/cameras/{id}/frame", tags=["Live Stream Monitoring"])
def get_frame_legacy(id: str):
if id not in manager.active_pipelines:
raise HTTPException(404)
def generate():
while id in manager.active_pipelines:
frame = manager.active_pipelines[id].latest_frame
if frame is not None:
_, buffer = cv2.imencode('.jpg', frame)
yield (b'--frame\r\nContent-Type: image/jpeg\r\n\r\n' + buffer.tobytes() + b'\r\n')
time.sleep(0.05)
return StreamingResponse(generate(), media_type="multipart/x-mixed-replace; boundary=frame")
# --- VIDEO PROCESSING ENDPOINTS ---
@app.post("/video/detect", tags=["Recorded Video Analysis"])
async def detect(background_tasks: BackgroundTasks, file: UploadFile = File(...)):
jid = str(uuid.uuid4())
in_p = os.path.join(UPLOAD_DIR, f"{jid}_{file.filename}")
out_p = os.path.join(OUTPUT_DIR, f"result_{jid}.mp4")
with open(in_p, "wb") as f: f.write(await file.read())
jobs[jid] = {"status": "processing", "progress": 0, "output_path": out_p, "stop_requested": False, "filename": file.filename}
# We use a lambda to run the async function in background tasks
background_tasks.add_task(lambda: asyncio.run(process_video_file(jid, in_p, out_p)))
return {"job_id": jid}
@app.get("/video/jobs", tags=["Recorded Video Analysis"])
async def list_jobs():
return [{"job_id": j, "status": d["status"], "progress": f"{d['progress']}%"} for j, d in jobs.items()]
@app.get("/video/status/{job_id}", tags=["Recorded Video Analysis"])
async def get_status(job_id: str):
return jobs.get(job_id, {"error": "Not found"})
@app.get("/video/download/{job_id}", tags=["Recorded Video Analysis"])
async def download(job_id: str):
if job_id in jobs and jobs[job_id]["status"] == "completed":
return FileResponse(jobs[job_id]["output_path"])
raise HTTPException(400, "File not ready")
if __name__ == "__main__":
import uvicorn
uvicorn.run(app, host="0.0.0.0", port=7860) |