app.py

import gradio as gr
import numpy as np
from PIL import Image, ImageDraw
import json
from typing import Tuple, List, Dict, Any
import time
import threading
import queue
import os
import pickle
from datetime import datetime
import hashlib

# --- Dependency Checks ---

# Try to import cv2
try:
    import cv2
    CV2_AVAILABLE = True
except ImportError:
    CV2_AVAILABLE = False
    print("Warning: OpenCV (cv2) not available. Using fallback image processing.")

# Try to import face recognition
try:
    import face_recognition
    FACE_REC_AVAILABLE = True
except ImportError:
    FACE_REC_AVAILABLE = False
    print("Warning: face_recognition not available. Using basic face detection only.")

# Try to import audio
try:
    import pyaudio
    AUDIO_AVAILABLE = True
except ImportError:
    AUDIO_AVAILABLE = False
    print("Warning: Audio libraries not available. Using silent alarm mode.")

# Try to import TensorFlow (Pet detection)
try:
    import tensorflow as tf
    from tensorflow.keras.applications.mobilenet_v2 import MobileNetV2, preprocess_input, decode_predictions
    TF_AVAILABLE = True
except ImportError:
    TF_AVAILABLE = False
    print("Warning: TensorFlow not available. Using basic pet detection only.")

# Try to import Ultralytics (YOLO)
try:
    from ultralytics import YOLO
    YOLO_AVAILABLE = True
except ImportError:
    YOLO_AVAILABLE = False
    print("Warning: Ultralytics (YOLO) not available. Using MobileNet/Simulation only.")

# --- Audio & Alarm System ---

def generate_tone(frequency, duration, sample_rate=44100, volume=0.5):
    """Generate a simple tone."""
    if not AUDIO_AVAILABLE:
        return None
    frames = int(duration * sample_rate)
    arr = np.zeros(frames)
    for i in range(frames):
        arr[i] = volume * np.sin(2 * np.pi * frequency * i / sample_rate)
    return arr.astype(np.float32)

def play_sound(sound_type):
    """Play different alarm sounds."""
    if not AUDIO_AVAILABLE:
        print(f"Alarm: {sound_type} (audio not available)")
        return
    
    p = pyaudio.PyAudio()
    sound_patterns = {
        "Beep": [(440, 0.2), (440, 0.2)],
        "Siren": [(600, 0.1), (800, 0.1), (600, 0.1), (800, 0.1)],
        "Chime": [(523, 0.3), (659, 0.3), (784, 0.5)],
        "Alert": [(1000, 0.1), (1500, 0.1), (2000, 0.1)],
        "Buzzer": [(200, 0.5)],
        "Ring": [(800, 0.2), (600, 0.2), (800, 0.2), (600, 0.2)],
        "Woof": [(600, 0.3), (400, 0.2)],
        "Meow": [(800, 0.2), (1000, 0.1)]
    }
    
    try:
        stream = p.open(format=pyaudio.paFloat32, channels=1, rate=44100, output=True)
        if sound_type in sound_patterns:
            for freq, duration in sound_patterns[sound_type]:
                tone = generate_tone(freq, duration)
                if tone is not None:
                    stream.write(tone.tobytes())
        stream.stop_stream()
        stream.close()
    except Exception as e:
        print(f"Error playing sound: {e}")
    finally:
        p.terminate()

class AlarmSystem:
    def __init__(self):
        self.alarm_queue = queue.Queue()
        self.alarm_thread = threading.Thread(target=self._alarm_worker, daemon=True)
        self.alarm_thread.start()
        self.last_alarm_time = 0
        self.alarm_cooldown = 2
    
    def _alarm_worker(self):
        while True:
            try:
                sound_type = self.alarm_queue.get(timeout=1)
                if sound_type:
                    play_sound(sound_type)
                self.alarm_queue.task_done()
            except queue.Empty:
                continue
            except Exception as e:
                print(f"Alarm worker error: {e}")
    
    def trigger_alarm(self, sound_type):
        current_time = time.time()
        if current_time - self.last_alarm_time > self.alarm_cooldown:
            self.alarm_queue.put(sound_type)
            self.last_alarm_time = current_time
            return True
        return False

# --- Databases ---

class PetDatabase:
    def __init__(self, db_file="pet_database.pkl"):
        self.db_file = db_file
        self.known_pets = {}
        self.load_database()
    
    def load_database(self):
        if os.path.exists(self.db_file):
            try:
                with open(self.db_file, 'rb') as f:
                    self.known_pets = pickle.load(f)
            except Exception as e:
                print(f"Error loading pet database: {e}")
                self.known_pets = {}
    
    def save_database(self):
        try:
            with open(self.db_file, 'wb') as f:
                pickle.dump(self.known_pets, f)
        except Exception as e:
            print(f"Error saving pet database: {e}")
    
    def add_pet(self, name, pet_type, breed, features, image_path=None):
        pet_id = hashlib.md5(f"{name}_{pet_type}_{datetime.now().isoformat()}".encode()).hexdigest()[:8]
        self.known_pets[pet_id] = {
            "name": name, "type": pet_type, "breed": breed,
            "features": features, "image_path": image_path,
            "registered_at": datetime.now().isoformat()
        }
        self.save_database()
        return pet_id
    
    def remove_pet(self, pet_id):
        if pet_id in self.known_pets:
            del self.known_pets[pet_id]
            self.save_database()
            return True
        return False
    
    def get_all_pets(self):
        return self.known_pets
    
    def recognize_pet(self, features, pet_type, tolerance=0.7):
        if not self.known_pets: return None, 0.0
        best_match, best_similarity = None, 0.0
        for pet_id, pet_data in self.known_pets.items():
            if pet_data["type"] == pet_type and pet_data["features"] is not None and features is not None:
                try:
                    similarity = np.dot(features, known_features) / (np.linalg.norm(features) * np.linalg.norm(known_features))
                    if similarity > tolerance and similarity > best_similarity:
                        best_similarity = similarity
                        best_match = pet_id
                except: continue
        return (best_match, best_similarity) if best_match else (None, 0.0)

class FaceDatabase:
    def __init__(self, db_file="face_database.pkl"):
        self.db_file = db_file
        self.known_faces = {}
        self.load_database()
    
    def load_database(self):
        if os.path.exists(self.db_file):
            try:
                with open(self.db_file, 'rb') as f:
                    self.known_faces = pickle.load(f)
            except Exception as e:
                self.known_faces = {}
    
    def save_database(self):
        try:
            with open(self.db_file, 'wb') as f:
                pickle.dump(self.known_faces, f)
        except Exception as e:
            print(f"Error saving face database: {e}")
    
    def add_face(self, name, encoding, image_path=None):
        face_id = hashlib.md5(f"{name}_{datetime.now().isoformat()}".encode()).hexdigest()[:8]
        self.known_faces[face_id] = {
            "name": name, "encoding": encoding, "image_path": image_path,
            "registered_at": datetime.now().isoformat()
        }
        self.save_database()
        return face_id
    
    def remove_face(self, face_id):
        if face_id in self.known_faces:
            del self.known_faces[face_id]
            self.save_database()
            return True
        return False
    
    def get_all_faces(self):
        return self.known_faces
    
    def recognize_face(self, face_encoding, tolerance=0.6):
        if not self.known_faces or not FACE_REC_AVAILABLE: return None, 0.0
        best_match, best_distance = None, float('inf')
        for face_id, face_data in self.known_faces.items():
            known_encoding = face_data["encoding"]
            if len(known_encoding) > 0 and len(face_encoding) > 0:
                try:
                    distance = face_recognition.face_distance([known_encoding], face_encoding)[0]
                    if distance < tolerance and distance < best_distance:
                        best_distance = distance
                        best_match = face_id
                except: continue
        return (best_match, 1.0 - best_distance) if best_match else (None, 0.0)

# --- Global Objects ---
alarm_system = AlarmSystem()
face_db = FaceDatabase()
pet_db = PetDatabase()

# --- Feature Extraction & Detection Logic ---

def extract_pet_features(image, pet_type):
    try:
        if isinstance(image, Image.Image): image = np.array(image)
        if CV2_AVAILABLE:
            image_resized = cv2.resize(image, (224, 224))
            image_rgb = cv2.cvtColor(image_resized, cv2.COLOR_BGR2RGB) if image_resized.shape[2] == 3 else image_resized
            features = (image_rgb.astype(np.float32) / 255.0).flatten()[::8]
            return features
        else:
            image_resized = np.array(Image.fromarray(image).resize((64, 64)))
            features = np.concatenate([np.histogram(image_resized[:,:,i], bins=16, range=(0,256))[0] for i in range(3)])
            return features.astype(np.float32)
    except: return None

def detect_dogs_and_cats(image, confidence):
    try:
        if isinstance(image, Image.Image): image = np.array(image)
        detections = []
        if TF_AVAILABLE:
            try:
                image_resized = cv2.resize(image, (224, 224))
                image_batch = np.expand_dims(image_resized, axis=0)
                image_preprocessed = preprocess_input(image_batch)
                if not hasattr(detect_dogs_and_cats, 'model'):
                    detect_dogs_and_cats.model = MobileNetV2(weights='imagenet')
                predictions = detect_dogs_and_cats.model.predict(image_preprocessed, verbose=0)
                decoded = decode_predictions(predictions, top=5)[0]
                
                dog_cat_classes = {
                    'golden_retriever': 'Dog (Golden Retriever)', 'Labrador_retriever': 'Dog (Labrador)',
                    'German_shepherd': 'Dog (German Shepherd)', 'beagle': 'Dog (Beagle)',
                    'Siamese_cat': 'Cat (Siamese)', 'Persian_cat': 'Cat (Persian)',
                    'tabby_cat': 'Cat (Tabby)', 'Egyptian_cat': 'Cat (Egyptian)'
                }
                for _, label, score in decoded:
                    for class_key, display_name in dog_cat_classes.items():
                        if class_key in label.lower() and score > confidence:
                            detections.append({
                                "bbox": [50, 50, image.shape[1]-100, image.shape[0]-100],
                                "confidence": round(float(score), 3),
                                "label": display_name, "type": "pet"
                            })
                            break
            except Exception as e: print(f"TensorFlow detection failed: {e}")
        
        if not detections: # Fallback simulation
            if np.random.random() > 0.3:
                pet_types = [{"label": "Dog", "type": "pet"}, {"label": "Cat", "type": "pet"}]
                pet_info = np.random.choice(pet_types)
                detections.append({
                    "bbox": [np.random.randint(0, max(1, image.shape[1]//2)), np.random.randint(0, max(1, image.shape[0]//2)), 200, 200],
                    "confidence": round(np.random.uniform(confidence, 0.95), 3),
                    "label": pet_info["label"], "type": pet_info["type"],
                    "recognized": None, "recognition_confidence": 0.0
                })
        return detections
    except: return []

def extract_face_encoding(image):
    if not FACE_REC_AVAILABLE: return None
    try:
        rgb_image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB) if isinstance(image, np.ndarray) else np.array(image)
        locs = face_recognition.face_locations(rgb_image)
        encs = face_recognition.face_encodings(rgb_image, locs)
        return encs[0] if encs else None
    except: return None

def register_face(image, name):
    if image is None or not name.strip(): return False, "Invalid input"
    encoding = extract_face_encoding(image)
    if encoding is None: return False, "No face detected"
    
    timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
    image_filename = f"registered_faces/{name.strip()}_{timestamp}.jpg"
    os.makedirs("registered_faces", exist_ok=True)
    Image.fromarray(image).save(image_filename)
    face_id = face_db.add_face(name.strip(), encoding, image_filename)
    return True, f"Registered face: {name} (ID: {face_id})"

def register_pet(image, name, pet_type, breed):
    if image is None or not name.strip(): return False, "Invalid input"
    features = extract_pet_features(image, pet_type)
    if features is None: return False, "Could not extract features"
    
    timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
    image_filename = f"registered_pets/{pet_type}_{name.strip()}_{timestamp}.jpg"
    os.makedirs("registered_pets", exist_ok=True)
    Image.fromarray(image).save(image_filename)
    pet_id = pet_db.add_pet(name.strip(), pet_type, breed, features, image_filename)
    return True, f"Registered {pet_type}: {name} (ID: {pet_id})"

def load_detection_models():
    face_cascade, object_net, object_classes = None, None, []
    if CV2_AVAILABLE:
        try:
            face_cascade = cv2.CascadeClassifier(cv2.data.haarcascades + 'haarcascade_frontalface_default.xml')
            try:
                object_net = cv2.dnn.readNetFromCaffe("MobileNetSSD_deploy.prototxt", "MobileNetSSD_deploy.caffemodel")
                object_classes = ["background", "aeroplane", "bicycle", "bird", "boat", "bottle", "bus", "car", "cat", "chair", "cow", "diningtable", "dog", "horse", "motorbike", "person", "pottedplant", "sheep", "sofa", "train", "tvmonitor"]
            except: pass
        except: pass
    return face_cascade, object_net, object_classes

face_cascade, object_net, object_classes = load_detection_models()

# --- Detection Implementations ---

def detect_faces_cv2(image, face_cascade, confidence):
    try:
        gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)
        faces = face_cascade.detectMultiScale(gray, 1.1, 5)
        results = []
        encodings = face_recognition.face_encodings(image) if FACE_REC_AVAILABLE else [None]*len(faces)
        
        for i, (x, y, w, h) in enumerate(faces):
            rec_name, rec_conf = None, 0.0
            if i < len(encodings) and encodings[i] is not None:
                fid, conf = face_db.recognize_face(encodings[i])
                if fid:
                    rec_name = face_db.known_faces[fid]["name"]
                    rec_conf = conf
            results.append({
                "bbox": [int(x), int(y), int(w), int(h)],
                "confidence": 0.9, "label": "face",
                "recognized": rec_name, "recognition_confidence": round(rec_conf, 3)
            })
        return results
    except: return []

def detect_objects_cv2(image, net, classes, confidence):
    try:
        if net is None: return []
        h, w = image.shape[:2]
        blob = cv2.dnn.blobFromImage(image, 0.007843, (300, 300), 127.5)
        net.setInput(blob)
        detections = net.forward()
        objects = []
        for i in range(detections.shape[2]):
            score = detections[0, 0, i, 2]
            if score > confidence:
                idx = int(detections[0, 0, i, 1])
                if idx < len(classes):
                    x1, y1 = int(detections[0, 0, i, 3] * w), int(detections[0, 0, i, 4] * h)
                    x2, y2 = int(detections[0, 0, i, 5] * w), int(detections[0, 0, i, 6] * h)
                    objects.append({
                        "bbox": [x1, y1, x2 - x1, y2 - y1],
                        "confidence": round(float(score), 3), "label": classes[idx]
                    })
        return objects
    except: return []

def detect_objects_yolo(image, model_name, confidence):
    """Detect objects using YOLOv8."""
    if not YOLO_AVAILABLE:
        return []
    try:
        # Load model (lazy loading)
        if not hasattr(detect_objects_yolo, 'models'):
            detect_objects_yolo.models = {}
        
        if model_name not in detect_objects_yolo.models:
            print(f"Loading YOLO model: {model_name}")
            detect_objects_yolo.models[model_name] = YOLO(f"{model_name}.pt")
            
        model = detect_objects_yolo.models[model_name]
        results = model(image, conf=confidence, verbose=False)
        
        objects = []
        for r in results:
            boxes = r.boxes
            for box in boxes:
                x1, y1, x2, y2 = box.xyxy[0].tolist()
                conf = float(box.conf[0])
                cls = int(box.cls[0])
                label = model.names[cls]
                
                objects.append({
                    "bbox": [int(x1), int(y1), int(x2-x1), int(y2-y1)],
                    "confidence": round(conf, 3),
                    "label": label
                })
        return objects
    except Exception as e:
        print(f"YOLO detection error: {e}")
        return []

def detect_objects_pil(image, confidence):
    # Simulation fallback
    width, height = Image.fromarray(image).size
    objects = []
    for i in range(min(5, np.random.randint(0, 5) + 1)):
        x, y = np.random.randint(0, width-100), np.random.randint(0, height-100)
        objects.append({
            "bbox": [x, y, 100, 100],
            "confidence": round(np.random.uniform(0.4, 0.9), 3),
            "label": np.random.choice(["person", "car", "bottle", "chair", "laptop"])
        })
    return objects

def process_image(image, face_cascade, object_net, object_classes, enable_face, enable_objects, enable_pets, face_conf, object_conf, pet_conf, object_model_type):
    face_results, object_results, pet_results = [], [], []
    
    if enable_face:
        face_results = detect_faces_cv2(image, face_cascade, face_conf) if CV2_AVAILABLE and face_cascade else []
    
    if enable_objects:
        if object_model_type.startswith("YOLO") and YOLO_AVAILABLE:
            # Map readable name to model filename
            yolo_map = {
                "YOLOv8 Nano": "yolov8n",
                "YOLOv8 Small": "yolov8s",
                "YOLOv8 Medium": "yolov8m"
            }
            model_file = yolo_map.get(object_model_type, "yolov8n")
            object_results = detect_objects_yolo(image, model_file, object_conf)
        elif object_model_type == "MobileNet SSD" and CV2_AVAILABLE and object_net:
            object_results = detect_objects_cv2(image, object_net, object_classes, object_conf)
        else:
            object_results = detect_objects_pil(image, object_conf)
    
    if enable_pets:
        pet_results = detect_dogs_and_cats(image, pet_conf)
        # Pet recognition logic
        for pet in pet_results:
            features = extract_pet_features(image, "dog" if "dog" in pet["label"].lower() else "cat")
            if features is not None:
                pid, conf = pet_db.recognize_pet(features, "dog" if "dog" in pet["label"].lower() else "cat")
                if pid:
                    pet["recognized"] = pet_db.known_pets[pid]["name"]
                    pet["recognition_confidence"] = round(conf, 3)
    
    return image.copy(), face_results, object_results, pet_results

def draw_detections(image, face_results, object_results, pet_results, show_labels, box_color):
    pil_image = Image.fromarray(image)
    draw = ImageDraw.Draw(pil_image)
    colors = {"red": (255,0,0), "green": (0,255,0), "blue": (0,0,255)}
    base_color = colors.get(box_color, (255,0,0))
    
    for results, default_color in [(face_results, base_color), (object_results, base_color), (pet_results, (255,140,0))]:
        for res in results:
            x, y, w, h = res["bbox"]
            color = (0,255,0) if res.get("recognized") else default_color
            draw.rectangle([x, y, x+w, y+h], outline=color, width=3)
            if show_labels:
                label = f"{res.get('recognized', res['label'])} {res.get('confidence', '')}"
                draw.text((x, y-20), label, fill=color)
    return np.array(pil_image)

def check_and_trigger_alarm(face_results, object_results, pet_results, alarm_settings):
    if not alarm_settings.get("alarm_enabled"): return False, "Alarm disabled"
    triggered, reason = False, ""
    
    if alarm_settings.get("face_alarm") and face_results:
        if alarm_settings.get("recognized_faces_only"):
            if any(f.get("recognized") for f in face_results): triggered, reason = True, "Recognized Face"
        else: triggered, reason = True, "Face Detected"
        
    elif alarm_settings.get("object_alarm") and object_results:
        targets = alarm_settings.get("target_objects", [])
        if targets:
            if any(obj["label"] in targets for obj in object_results): triggered, reason = True, "Target Object"
        else: triggered, reason = True, "Object Detected"
        
    elif alarm_settings.get("pet_alarm") and pet_results:
        triggered, reason = True, "Pet Detected"
        
    if triggered:
        sound = "Woof" if pet_results and alarm_settings.get("pet_sound_enabled") else alarm_settings.get("alarm_sound", "Beep")
        if alarm_system.trigger_alarm(sound): return True, f"🚨 ALARM: {reason}"
        return False, "Alarm Cooldown"
    return False, "No conditions met"

def process_pipeline(image, enable_face, enable_obj, enable_pet, conf_face, conf_obj, conf_pet, draw, labels, color, obj_model, alarm_settings):
    if image is None: return None, "", "", "", ""
    if isinstance(image, Image.Image): image = np.array(image)
    
    proc_img, faces, objects, pets = process_image(image, face_cascade, object_net, object_classes, enable_face, enable_obj, enable_pet, conf_face, conf_obj, conf_pet, obj_model)
    
    _, alarm_msg = check_and_trigger_alarm(faces, objects, pets, alarm_settings)
    
    if draw:
        proc_img = draw_detections(proc_img, faces, objects, pets, labels, color)
        
    return proc_img, json.dumps(faces, indent=2), json.dumps(objects, indent=2), json.dumps(pets, indent=2), alarm_msg

def get_detection_statistics():
    stats = {
        "face_detection": {"model": "Haar Cascade" if CV2_AVAILABLE else "Simulated", "registered": len(face_db.known_faces)},
        "object_detection": {
            "available_models": ["MobileNet SSD", "YOLOv8 Nano", "YOLOv8 Small"] if YOLO_AVAILABLE else ["MobileNet SSD", "Simulated"],
            "ultralytics_installed": YOLO_AVAILABLE,
            "opencv_installed": CV2_AVAILABLE
        },
        "pet_detection": {"model": "MobileNetV2" if TF_AVAILABLE else "Simulated", "registered": len(pet_db.known_pets)}
    }
    return json.dumps(stats, indent=2)

# --- Gradio UI ---

custom_css = """
.warning-box { background-color: #fff3cd; border: 1px solid #ffeaa7; padding: 10px; border-radius: 5px; }
.alarm-box { background-color: #f8d7da; border: 1px solid #f5c6cb; padding: 10px; border-radius: 5px; animation: pulse 1s infinite; }
@keyframes pulse { 0% { opacity: 1; } 50% { opacity: 0.7; } 100% { opacity: 1; } }
"""

with gr.Blocks(css=custom_css, title="Security & Recognition System") as demo:
    gr.Markdown("""
    # 🔍 <a href="https://huggingface.co/spaces/akhaliq/anycoder" target="_blank">Built with anycoder</a>: Multi-Model Recognition Platform
    Real-time detection with support for **MobileNet SSD** and **YOLOv8** models.
    """)
    
    # Warnings
    if not CV2_AVAILABLE: gr.HTML('<div class="warning-box">⚠️ OpenCV not available. Using Simulation.</div>')
    if not YOLO_AVAILABLE: gr.HTML('<div class="warning-box">⚠️ Ultralytics not available. YOLO models disabled.</div>')
    
    with gr.Tabs():
        with gr.TabItem("🔍 Live Detection"):
            with gr.Row():
                with gr.Column(scale=2):
                    input_image = gr.Image(label="Input Feed", sources=["webcam", "upload"], type="numpy")
                    with gr.Row():
                        enable_face = gr.Checkbox(label="Face Detection", value=True)
                        enable_obj = gr.Checkbox(label="Object Detection", value=True)
                        enable_pet = gr.Checkbox(label="Pet Detection", value=True)
                    
                    # NEW: Object Model Selector
                    obj_model_selector = gr.Dropdown(
                        label="Object Detection Model",
                        choices=["MobileNet SSD", "YOLOv8 Nano", "YOLOv8 Small", "YOLOv8 Medium", "Simulation"],
                        value="YOLOv8 Nano" if YOLO_AVAILABLE else "MobileNet SSD",
                        interactive=True,
                        info="Select the architecture used for object detection."
                    )
                    
                    with gr.Accordion("Confidence Thresholds", open=False):
                        conf_face = gr.Slider(0, 1, 0.5, label="Face Confidence")
                        conf_obj = gr.Slider(0, 1, 0.5, label="Object Confidence")
                        conf_pet = gr.Slider(0, 1, 0.5, label="Pet Confidence")
                        
                    with gr.Accordion("Visualization Settings", open=False):
                        draw_boxes = gr.Checkbox(label="Draw Boxes", value=True)
                        show_labels = gr.Checkbox(label="Show Labels", value=True)
                        box_color = gr.Dropdown(choices=["red", "green", "blue"], value="red", label="Box Color")

                    detect_btn = gr.Button("Analyze Frame", variant="primary")
                    
                with gr.Column(scale=2):
                    output_image = gr.Image(label="Processed Output")
                    alarm_display = gr.Textbox(label="Alarm Status")
                    with gr.Accordion("Raw Data", open=False):
                        json_face = gr.JSON(label="Face Data")
                        json_obj = gr.JSON(label="Object Data")
                        json_pet = gr.JSON(label="Pet Data")

        with gr.TabItem("⚙️ Alarm Settings"):
            alarm_enabled = gr.Checkbox(label="Master Alarm Switch", value=False)
            with gr.Row():
                face_alarm = gr.Checkbox(label="Trigger on Face")
                obj_alarm = gr.Checkbox(label="Trigger on Object")
                pet_alarm = gr.Checkbox(label="Trigger on Pet")
            
            target_objs = gr.Dropdown(
                choices=["person", "car", "cell phone", "laptop", "bottle", "chair"], 
                multiselect=True, label="Target Objects for Alarm"
            )
            alarm_sound = gr.Dropdown(choices=["Beep", "Siren", "Woof", "Meow"], value="Beep", label="Sound Type")
            
            # Hidden state to store settings dict
            alarm_settings_state = gr.State({
                "alarm_enabled": False, "face_alarm": False, "object_alarm": False, 
                "pet_alarm": False, "target_objects": [], "alarm_sound": "Beep"
            })
            
            def update_settings(enabled, f, o, p, targets, sound):
                return {
                    "alarm_enabled": enabled, "face_alarm": f, "object_alarm": o, 
                    "pet_alarm": p, "target_objects": targets, "alarm_sound": sound,
                    "recognized_faces_only": False, "pet_sound_enabled": True
                }
            
            settings_inputs = [alarm_enabled, face_alarm, obj_alarm, pet_alarm, target_objs, alarm_sound]
            for inp in settings_inputs:
                inp.change(update_settings, settings_inputs, alarm_settings_state)

        with gr.TabItem("📝 Registration"):
            with gr.Row():
                with gr.Column():
                    gr.Markdown("### Face Registration")
                    reg_face_img = gr.Image(sources=["webcam", "upload"], type="numpy", height=300)
                    reg_face_name = gr.Textbox(label="Name")
                    reg_face_btn = gr.Button("Register Face")
                    reg_face_out = gr.Textbox(label="Result")
                    reg_face_btn.click(register_face, [reg_face_img, reg_face_name], reg_face_out)
                
                with gr.Column():
                    gr.Markdown("### Pet Registration")
                    reg_pet_img = gr.Image(sources=["webcam", "upload"], type="numpy", height=300)
                    reg_pet_name = gr.Textbox(label="Pet Name")
                    reg_pet_type = gr.Dropdown(["dog", "cat"], label="Type")
                    reg_pet_breed = gr.Textbox(label="Breed")
                    reg_pet_btn = gr.Button("Register Pet")
                    reg_pet_out = gr.Textbox(label="Result")
                    reg_pet_btn.click(register_pet, [reg_pet_img, reg_pet_name, reg_pet_type, reg_pet_breed], reg_pet_out)

        with gr.TabItem("📊 System Info"):
            refresh_stats = gr.Button("Refresh Statistics")
            stats_json = gr.JSON(label="System Statistics")
            refresh_stats.click(get_detection_statistics, None, stats_json)

    # Main detection event
    detect_btn.click(
        process_