# app.py import os import gradio as gr from PIL import Image import torch import numpy as np import cv2 from transformers import ( CLIPProcessor, CLIPModel, AutoProcessor ) import time import logging # Setup logging for continuous feedback logging.basicConfig(level=logging.INFO) logger = logging.getLogger(__name__) # ————————————————————————————— # CONFIG: set your HF token here or via env var HF_TOKEN HF_TOKEN = os.getenv("HF_TOKEN") device = torch.device("cuda" if torch.cuda.is_available() else "cpu") # 1. CLIP for breed, age, basic health clip_model = CLIPModel.from_pretrained( "openai/clip-vit-base-patch16", token=HF_TOKEN ).to(device) clip_processor = CLIPProcessor.from_pretrained( "openai/clip-vit-base-patch16", token=HF_TOKEN ) # 2. Alternative medical analysis model try: medical_processor = AutoProcessor.from_pretrained( "microsoft/BiomedCLIP-PubMedBERT_256-vit_base_patch16_224", token=HF_TOKEN ) medical_model = CLIPModel.from_pretrained( "microsoft/BiomedCLIP-PubMedBERT_256-vit_base_patch16_224", token=HF_TOKEN ).to(device) MEDICAL_MODEL_AVAILABLE = True except: medical_processor = clip_processor medical_model = clip_model MEDICAL_MODEL_AVAILABLE = False # Stanford Dogs & lifespans STANFORD_BREEDS = [ "afghan hound", "african hunting dog", "airedale", "american staffordshire terrier", "appenzeller", "australian terrier", "basenji", "basset", "beagle", "bedlington terrier", "bernese mountain dog", "black-and-tan coonhound", "blenheim spaniel", "bloodhound", "bluetick", "border collie", "border terrier", "borzoi", "boston bull", "bouvier des flandres", "boxer", "brabancon griffon", "briard", "brittany spaniel", "bull mastiff", "cairn", "cardigan", "chesapeake bay retriever", "chihuahua", "chow", "clumber", "cocker spaniel", "collie", "curly-coated retriever", "dandie dinmont", "dhole", "dingo", "doberman", "english foxhound", "english setter", "english springer", "entlebucher", "eskimo dog", "flat-coated retriever", "french bulldog", "german shepherd", "german short-haired pointer", "giant schnauzer", "golden retriever", "gordon setter", "great dane", "great pyrenees", "greater swiss mountain dog", "groenendael", "ibizan hound", "irish setter", "irish terrier", "irish water spaniel", "irish wolfhound", "italian greyhound", "japanese spaniel", "keeshond", "kelpie", "kerry blue terrier", "komondor", "kuvasz", "labrador retriever", "lakeland terrier", "leonberg", "lhasa", "malamute", "malinois", "maltese dog", "mexican hairless", "miniature pinscher", "miniature poodle", "miniature schnauzer", "newfoundland", "norfolk terrier", "norwegian elkhound", "norwich terrier", "old english sheepdog", "otterhound", "papillon", "pekinese", "pembroke", "pomeranian", "pug", "redbone", "rhodesian ridgeback", "rottweiler", "saint bernard", "saluki", "samoyed", "schipperke", "scotch terrier", "scottish deerhound", "sealyham terrier", "shetland sheepdog", "shih tzu", "siberian husky", "silky terrier", "soft-coated wheaten terrier", "staffordshire bullterrier", "standard poodle", "standard schnauzer", "sussex spaniel", "tibetan mastiff", "tibetan terrier", "toy poodle", "toy terrier", "vizsla", "walker hound", "weimaraner", "welsh springer spaniel", "west highland white terrier", "whippet", "wire-haired fox terrier", "yorkshire terrier" ] BREED_LIFESPAN = { "afghan hound": 11.1, "african hunting dog": 10.5, "airedale": 11.5, "american staffordshire terrier": 12.5, "appenzeller": 13.0, "australian terrier": 13.5, "basenji": 12.1, "basset": 12.5, "beagle": 12.5, "bedlington terrier": 13.7, "bernese mountain dog": 10.1, "black-and-tan coonhound": 10.8, "blenheim spaniel": 13.3, "bloodhound": 9.3, "bluetick": 11.0, "border collie": 13.1, "border terrier": 14.2, "borzoi": 12.0, "boston bull": 11.8, "bouvier des flandres": 11.3, "boxer": 11.3, "brabancon griffon": 13.0, "briard": 12.6, "brittany spaniel": 13.5, "bull mastiff": 10.2, "cairn": 14.0, "cardigan": 13.2, "chesapeake bay retriever": 11.6, "chihuahua": 11.8, "chow": 12.1, "clumber": 12.3, "cocker spaniel": 13.3, "collie": 13.3, "curly-coated retriever": 12.2, "dandie dinmont": 12.8, "dhole": 10.0, "dingo": 10.0, "doberman": 11.2, "english foxhound": 13.0, "english setter": 13.1, "english springer": 13.5, "entlebucher": 13.0, "eskimo dog": 11.3, "flat-coated retriever": 11.7, "french bulldog": 9.8, "german shepherd": 11.3, "german short-haired pointer": 13.4, "giant schnauzer": 12.1, "golden retriever": 13.2, "gordon setter": 12.4, "great dane": 10.6, "great pyrenees": 10.9, "greater swiss mountain dog": 10.9, "groenendael": 12.0, "ibizan hound": 13.3, "irish setter": 12.9, "irish terrier": 13.5, "irish water spaniel": 10.8, "irish wolfhound": 9.9, "italian greyhound": 14.0, "japanese spaniel": 13.3, "keeshond": 12.3, "kelpie": 12.0, "kerry blue terrier": 12.4, "komondor": 10.5, "kuvasz": 10.5, "labrador retriever": 13.1, "lakeland terrier": 14.2, "leonberg": 10.0, "lhasa": 14.0, "malamute": 11.3, "malinois": 12.0, "maltese dog": 13.1, "mexican hairless": 13.0, "miniature pinscher": 13.7, "miniature poodle": 14.0, "miniature schnauzer": 13.3, "newfoundland": 11.0, "norfolk terrier": 13.5, "norwegian elkhound": 13.0, "norwich terrier": 14.0, "old english sheepdog": 12.1, "otterhound": 12.0, "papillon": 14.5, "pekinese": 13.3, "pembroke": 13.2, "pomeranian": 12.2, "pug": 11.6, "redbone": 12.0, "rhodesian ridgeback": 12.0, "rottweiler": 10.6, "saint bernard": 9.3, "saluki": 13.3, "samoyed": 13.1, "schipperke": 14.2, "scotch terrier": 12.7, "scottish deerhound": 10.5, "sealyham terrier": 13.1, "shetland sheepdog": 13.4, "shih tzu": 12.8, "siberian husky": 11.9, "silky terrier": 13.3, "soft-coated wheaten terrier": 13.7, "staffordshire bullterrier": 12.0, "standard poodle": 14.0, "standard schnauzer": 13.0, "sussex spaniel": 13.5, "tibetan mastiff": 13.3, "tibetan terrier": 13.8, "toy poodle": 14.0, "toy terrier": 13.0, "vizsla": 13.5, "walker hound": 12.0, "weimaraner": 12.8, "welsh springer spaniel": 14.0, "west highland white terrier": 13.4, "whippet": 13.4, "wire-haired fox terrier": 13.5, "yorkshire terrier": 13.3 } # SHORTENED HRQOL Questionnaire HRQOL_QUESTIONNAIRE = { "vitality": { "title": "🔋 Vitality & Energy Assessment", "description": "Evaluate your dog's overall energy and responsiveness", "questions": [ { "id": "vitality_comprehensive", "text": "How would you rate your dog's overall vitality considering energy levels, playfulness, and responsiveness to exciting activities?", "options": [ "Excellent - Very energetic, always seeks play, immediate enthusiastic responses", "Very Good - Generally energetic, often seeks play, quick positive responses", "Good - Moderate energy, sometimes seeks play, moderate response time", "Fair - Lower energy, rarely seeks play, slow or delayed responses", "Poor - Very low energy, never seeks play, no response or negative reactions" ] } ], "weight": 0.25 }, "comfort": { "title": "😌 Comfort & Pain Management", "description": "Assess overall comfort and mobility", "questions": [ { "id": "comfort_comprehensive", "text": "How would you assess your dog's overall comfort considering activity comfort, pain frequency, and impact on daily life?", "options": [ "Excellent - Completely comfortable in all activities, never shows pain, no impact on daily life", "Very Good - Mostly comfortable with minor adjustments, rarely shows pain, minimal impact", "Good - Some discomfort in certain activities, occasional pain signs, moderate activity modifications", "Fair - Frequently uncomfortable, often shows pain, significant activity limitations", "Poor - Severe discomfort in most activities, daily pain signs, major activity restrictions" ] } ], "weight": 0.25 }, "emotional_wellbeing": { "title": "😊 Emotional Wellbeing", "description": "Evaluate mood, stress levels, and social engagement", "questions": [ { "id": "emotional_comprehensive", "text": "How would you describe your dog's overall emotional state considering mood, stress/anxiety levels, and family engagement?", "options": [ "Excellent - Very positive mood, never shows stress, highly engaged with family activities", "Very Good - Mostly positive mood, rarely shows stress, well engaged with family", "Good - Generally neutral mood, sometimes shows stress, moderately engaged when invited", "Fair - Often subdued mood, frequently shows stress, minimally engaged with encouragement", "Poor - Negative/depressed mood, constantly stressed, avoids family activities" ] } ], "weight": 0.25 }, "alertness": { "title": "🧠 Alertness & Cognition", "description": "Assess cognitive function and awareness", "questions": [ { "id": "alertness_comprehensive", "text": "How would you rate your dog's overall cognitive function considering awareness, command response, and focus during activities?", "options": [ "Excellent - Highly alert and aware, responds immediately to commands, maintains excellent focus", "Very Good - Alert and notices things quickly, usually responds quickly, good focus with occasional distraction", "Good - Moderately alert with some delay, sometimes needs repetition, moderate focus with difficulty concentrating", "Fair - Slightly alert and slow to notice, often needs multiple attempts, poor focus and easily distracted", "Poor - Not alert or confused, rarely responds to commands, cannot maintain attention or focus" ] } ], "weight": 0.25 } } # ====== ENHANCED BIOLOGICAL AGE PREDICTION FUNCTIONS ====== def predict_biological_age_enhanced(img: Image.Image, video_path: str, breed: str, hrqol_scores: dict, age: int = None): """Enhanced biological age prediction with accurate multi-factor analysis""" try: # 1. Base prediction using breed-specific aging curves breed_lifespan = BREED_LIFESPAN.get(breed.lower(), 12.0) # 2. Enhanced visual health indicators with detailed analysis health_indicators = analyze_health_indicators_detailed(img) # 3. HRQOL-based age adjustment with refined weighting hrqol_adjustment = calculate_hrqol_age_factor_refined(hrqol_scores) # 4. Video gait analysis (if available) gait_adjustment = 0 if video_path: gait_features = analyze_video_for_age_indicators_enhanced(video_path) gait_adjustment = gait_features.get('age_factor', 0) # 5. Multi-factor biological age calculation if age and age > 0: # Start with chronological age as base base_age = float(age) # Apply visual health assessment (stronger influence) visual_factor = health_indicators.get('age_factor', 0.0) * 0.8 # Apply HRQOL health adjustment (moderate influence) hrqol_factor = hrqol_adjustment * 0.6 # Apply gait/movement adjustment (if available) gait_factor = gait_adjustment * 0.4 if video_path else 0.0 # Calculate combined health impact total_health_impact = visual_factor + hrqol_factor + gait_factor # Apply health impact to biological age calculation # Positive factors = accelerated aging, Negative factors = slower aging biological_age = base_age * (1.0 + total_health_impact) # Add breed-specific aging rate adjustments breed_aging_rate = calculate_breed_aging_rate(breed, age, breed_lifespan) biological_age = biological_age * breed_aging_rate else: # When no chronological age provided, estimate from visual cues visual_age_estimate = estimate_age_from_visual_cues_enhanced(img, breed) # Apply health adjustments to visual estimate health_adjustment = (hrqol_adjustment + gait_adjustment) * 0.5 biological_age = visual_age_estimate * (1.0 + health_adjustment) # 6. Apply realistic constraints min_age = max(0.3, age * 0.7) if age else 0.3 max_age = min(breed_lifespan * 1.4, age * 1.6) if age else breed_lifespan * 1.2 biological_age = max(min_age, min(max_age, biological_age)) # 7. Calculate confidence and uncertainty prediction_confidence = calculate_prediction_confidence_enhanced(health_indicators, hrqol_scores, video_path, age) uncertainty = max(0.1, (1.0 - prediction_confidence) * 2.0) return { 'biological_age': round(biological_age, 1), 'uncertainty': round(uncertainty, 1), 'high_uncertainty': uncertainty > 1.5, 'vision_quality': compute_vision_quality_enhanced(img), 'breed_lifespan': breed_lifespan, 'confidence_factors': { 'visual_health': health_indicators, 'hrqol_factor': hrqol_adjustment, 'gait_factor': gait_adjustment, 'total_health_impact': total_health_impact if age else 0.0, 'prediction_confidence': prediction_confidence } } except Exception as e: logger.error(f"Error in enhanced age prediction: {e}") # Fallback calculation fallback_age = age * 1.1 if age else breed_lifespan * 0.4 return { 'biological_age': round(fallback_age, 1), 'uncertainty': 2.0, 'high_uncertainty': True, 'vision_quality': 0.5, 'breed_lifespan': breed_lifespan } def analyze_health_indicators_detailed(img: Image.Image): """Enhanced visual health analysis with detailed aging assessment""" try: # More comprehensive aging assessment prompts aging_prompts = [ "very young healthy puppy with baby features and perfect health", "young adult dog with excellent health and prime condition", "healthy mature adult dog with minor aging signs", "middle-aged dog with moderate aging and some health decline", "senior dog with visible aging and health deterioration", "elderly dog with significant aging and multiple health issues" ] inputs = clip_processor(text=aging_prompts, images=img, return_tensors="pt", padding=True).to(device) with torch.no_grad(): logits = clip_model(**inputs).logits_per_image.softmax(-1)[0].cpu().numpy() # More nuanced age factor weights (higher range for better distinction) age_weights = [-0.6, -0.3, -0.1, 0.2, 0.4, 0.7] age_factor = np.dot(logits, age_weights) # Additional physical condition analysis condition_prompts = [ "dog with excellent physical condition and youthful appearance", "dog with good physical condition and minimal aging", "dog with fair physical condition and moderate aging", "dog with poor physical condition and advanced aging" ] inputs2 = clip_processor(text=condition_prompts, images=img, return_tensors="pt", padding=True).to(device) with torch.no_grad(): condition_logits = clip_model(**inputs2).logits_per_image.softmax(-1)[0].cpu().numpy() condition_weights = [-0.4, -0.1, 0.2, 0.5] condition_factor = np.dot(condition_logits, condition_weights) # Combine factors with weighted importance combined_factor = (age_factor * 0.7) + (condition_factor * 0.3) return { 'age_factor': float(combined_factor), 'confidence': float(np.max(logits)), 'distribution': logits.tolist(), 'condition_factor': float(condition_factor) } except Exception as e: logger.error(f"Error in detailed health indicator analysis: {e}") return {'age_factor': 0.0, 'confidence': 0.5, 'distribution': [0.16]*6, 'condition_factor': 0.0} def calculate_hrqol_age_factor_refined(hrqol_scores: dict): """Refined HRQOL aging factor with stronger impact""" try: # Calculate weighted average with domain-specific importance domain_weights = { 'vitality': 0.35, # Highest correlation with aging 'comfort': 0.30, # Strong correlation with aging 'alertness': 0.25, # Cognitive aging indicator 'emotional_wellbeing': 0.10 # Secondary factor } weighted_score = sum( hrqol_scores.get(domain, 50) * weight for domain, weight in domain_weights.items() ) # More pronounced age factor calculation for better distinction if weighted_score >= 85: # Excellent health age_factor = -0.25 # Significantly slower aging elif weighted_score >= 70: # Very good health age_factor = -0.15 # Slower aging elif weighted_score >= 55: # Good health age_factor = -0.05 # Slightly slower aging elif weighted_score >= 40: # Fair health age_factor = 0.1 # Slightly accelerated aging elif weighted_score >= 25: # Poor health age_factor = 0.3 # Accelerated aging else: # Very poor health age_factor = 0.5 # Significantly accelerated aging return age_factor except Exception as e: logger.error(f"Error in refined HRQOL age factor calculation: {e}") return 0.0 def analyze_video_for_age_indicators_enhanced(video_path: str): """Enhanced video analysis with detailed movement assessment""" try: cap = cv2.VideoCapture(video_path) total_frames = int(cap.get(cv2.CAP_PROP_FRAME_COUNT)) if total_frames == 0: cap.release() return {'age_factor': 0.0} movement_scores = [] energy_scores = [] # Sample more frames for better accuracy frame_indices = np.linspace(0, total_frames-1, min(20, total_frames), dtype=int) for idx in frame_indices: cap.set(cv2.CAP_PROP_POS_FRAMES, idx) ret, frame = cap.read() if not ret: continue img = Image.fromarray(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)) # Enhanced movement assessment movement_prompts = [ "young dog with bouncy energetic playful movement", "adult dog with smooth confident coordinated movement", "older dog with careful measured slower movement", "senior dog with stiff labored difficult movement" ] inputs = clip_processor(text=movement_prompts, images=img, return_tensors="pt", padding=True).to(device) with torch.no_grad(): logits = clip_model(**inputs).logits_per_image.softmax(-1)[0].cpu().numpy() # Stronger movement scoring for better age distinction movement_weights = [-0.4, -0.1, 0.2, 0.5] movement_score = np.dot(logits, movement_weights) movement_scores.append(movement_score) # Energy level assessment energy_prompts = [ "very high energy enthusiastic dog", "good energy alert dog", "moderate energy calm dog", "low energy lethargic dog" ] inputs2 = clip_processor(text=energy_prompts, images=img, return_tensors="pt", padding=True).to(device) with torch.no_grad(): energy_logits = clip_model(**inputs2).logits_per_image.softmax(-1)[0].cpu().numpy() energy_weights = [-0.3, -0.1, 0.1, 0.4] energy_score = np.dot(energy_logits, energy_weights) energy_scores.append(energy_score) cap.release() if movement_scores and energy_scores: # Combine movement and energy with weighted importance avg_movement = np.mean(movement_scores) avg_energy = np.mean(energy_scores) combined_factor = (avg_movement * 0.7) + (avg_energy * 0.3) return { 'age_factor': float(combined_factor), 'movement_score': float(avg_movement), 'energy_score': float(avg_energy), 'sample_count': len(movement_scores) } else: return {'age_factor': 0.0} except Exception as e: logger.error(f"Error in enhanced video age analysis: {e}") return {'age_factor': 0.0} def calculate_breed_aging_rate(breed: str, current_age: int, breed_lifespan: float): """Calculate breed-specific aging rate adjustment""" try: # Calculate relative age within breed lifespan relative_age = current_age / breed_lifespan # Aging rate adjustments based on breed characteristics if relative_age < 0.2: # Very young (0-20% of lifespan) aging_rate = 0.95 # Slightly slower development elif relative_age < 0.4: # Young adult (20-40% of lifespan) aging_rate = 1.0 # Normal aging elif relative_age < 0.6: # Mature adult (40-60% of lifespan) aging_rate = 1.05 # Slightly accelerated elif relative_age < 0.8: # Senior (60-80% of lifespan) aging_rate = 1.15 # Accelerated aging else: # Elderly (80%+ of lifespan) aging_rate = 1.25 # Significantly accelerated # Breed-specific adjustments large_breeds = ["great dane", "saint bernard", "mastiff", "irish wolfhound"] small_breeds = ["chihuahua", "toy poodle", "papillon", "maltese dog"] if any(large_breed in breed.lower() for large_breed in large_breeds): aging_rate *= 1.1 # Large breeds age faster elif any(small_breed in breed.lower() for small_breed in small_breeds): aging_rate *= 0.95 # Small breeds age slower return aging_rate except Exception as e: logger.error(f"Error in breed aging rate calculation: {e}") return 1.0 def estimate_age_from_visual_cues_enhanced(img: Image.Image, breed: str): """Enhanced age estimation with more detailed visual analysis""" try: breed_lifespan = BREED_LIFESPAN.get(breed.lower(), 12.0) # More detailed age-specific descriptions age_ranges = [ (0.3, f"very young {breed} puppy with baby features and soft coat"), (1.0, f"young {breed} puppy with developing adult features"), (2.5, f"adolescent {breed} with youthful energy and developing body"), (4.0, f"young adult {breed} in peak physical condition"), (7.0, f"mature adult {breed} with full development and strength"), (10.0, f"middle-aged {breed} with some aging signs and experience"), (breed_lifespan * 0.85, f"senior {breed} with clear aging and wisdom"), (breed_lifespan, f"elderly {breed} with advanced aging and slower movement") ] age_prompts = [desc for _, desc in age_ranges] inputs = clip_processor(text=age_prompts, images=img, return_tensors="pt", padding=True).to(device) with torch.no_grad(): logits = clip_model(**inputs).logits_per_image.softmax(-1)[0].cpu().numpy() # Calculate weighted average age with higher precision ages = [age for age, _ in age_ranges] estimated_age = np.dot(logits, ages) # Apply confidence-based adjustment confidence = np.max(logits) if confidence < 0.4: # Low confidence # Default to middle age estimate estimated_age = breed_lifespan * 0.5 return max(0.2, min(breed_lifespan * 1.2, estimated_age)) except Exception as e: logger.error(f"Error in enhanced visual age estimation: {e}") return BREED_LIFESPAN.get(breed.lower(), 12.0) * 0.5 def calculate_prediction_confidence_enhanced(health_indicators: dict, hrqol_scores: dict, video_path: str, age: int = None): """Calculate enhanced prediction confidence""" try: confidence_factors = [] # Visual analysis confidence (higher weight) visual_conf = health_indicators.get('confidence', 0.5) confidence_factors.append(visual_conf * 0.4) # Chronological age availability (high importance) age_conf = 0.95 if age else 0.2 confidence_factors.append(age_conf * 0.3) # HRQOL completeness and consistency completed_domains = sum(1 for score in hrqol_scores.values() if score > 0) hrqol_conf = completed_domains / 4.0 confidence_factors.append(hrqol_conf * 0.2) # Video availability video_conf = 0.9 if video_path else 0.5 confidence_factors.append(video_conf * 0.1) overall_confidence = sum(confidence_factors) return min(1.0, overall_confidence) except Exception as e: return 0.5 def compute_vision_quality_enhanced(img: Image.Image): """Enhanced vision quality computation""" try: gray = cv2.cvtColor(np.array(img), cv2.COLOR_RGB2GRAY) # Sharpness calculation sharpness = cv2.Laplacian(gray, cv2.CV_64F).var() # Exposure quality mean_intensity = np.mean(gray) exposure_quality = 1.0 - abs(mean_intensity - 127.5) / 127.5 # Contrast quality contrast = np.std(gray) / 128.0 contrast_quality = min(1.0, contrast) # Combined quality score quality = (sharpness / 1500.0 * 0.5 + exposure_quality * 0.3 + contrast_quality * 0.2) quality = min(1.0, quality) return max(0.1, quality) except Exception as e: logger.error(f"Error in enhanced quality computation: {e}") return 0.5 # ====== EXISTING SUPPORT FUNCTIONS ====== def analyze_medical_image(img: Image.Image): health_conditions = [ "healthy normal dog", "dog with visible health issues", "dog showing signs of illness", "dog with poor body condition", "dog with excellent health" ] if MEDICAL_MODEL_AVAILABLE: inputs = medical_processor(text=health_conditions, images=img, return_tensors="pt", padding=True).to(device) with torch.no_grad(): logits = medical_model(**inputs).logits_per_image.softmax(-1)[0].cpu().numpy() else: inputs = clip_processor(text=health_conditions, images=img, return_tensors="pt", padding=True).to(device) with torch.no_grad(): logits = clip_model(**inputs).logits_per_image.softmax(-1)[0].cpu().numpy() idx = int(np.argmax(logits)) label = health_conditions[idx] conf = float(logits[idx]) return label, conf def classify_breed_and_health(img: Image.Image, override=None): inp = clip_processor(images=img, return_tensors="pt").to(device) with torch.no_grad(): feats = clip_model.get_image_features(**inp) text_prompts = [f"a photo of a {b}" for b in STANFORD_BREEDS] ti = clip_processor(text=text_prompts, return_tensors="pt", padding=True).to(device) with torch.no_grad(): tf = clip_model.get_text_features(**ti) sims = (feats @ tf.T).softmax(-1)[0].cpu().numpy() idx = int(np.argmax(sims)) breed = override or STANFORD_BREEDS[idx] breed_conf = float(sims[idx]) aspects = { "Coat Quality": ("shiny healthy coat","dull patchy fur"), "Eye Clarity": ("bright clear eyes","cloudy milky eyes"), "Body Condition": ("ideal muscle tone","visible ribs or bones"), "Dental Health": ("clean white teeth","yellow stained teeth") } health = {} for name,(p,n) in aspects.items(): ti = clip_processor(text=[p,n], return_tensors="pt", padding=True).to(device) with torch.no_grad(): tf2 = clip_model.get_text_features(**ti) sim2 = (feats @ tf2.T).softmax(-1)[0].cpu().numpy() choice = p if sim2[0]>sim2[1] else n health[name] = {"assessment":choice,"confidence":float(max(sim2))} return breed, breed_conf, health def analyze_video_gait(video_path): if not video_path: return None try: cap = cv2.VideoCapture(video_path) fps = cap.get(cv2.CAP_PROP_FPS) or 24 total = int(cap.get(cv2.CAP_PROP_FRAME_COUNT)) if total == 0: cap.release() return None indices = np.linspace(0, total-1, min(15, total), dtype=int) health_scores = [] movement_scores = [] vitality_scores = [] for i in indices: cap.set(cv2.CAP_PROP_POS_FRAMES, i) ret, frame = cap.read() if not ret: continue img = Image.fromarray(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)) # Health assessment _, health_conf = analyze_medical_image(img) health_scores.append(health_conf) # Movement assessment movement_prompts = ["dog moving normally", "dog limping or showing pain", "dog moving stiffly"] inputs = clip_processor(text=movement_prompts, images=img, return_tensors="pt", padding=True).to(device) with torch.no_grad(): movement_logits = clip_model(**inputs).logits_per_image.softmax(-1)[0].cpu().numpy() movement_scores.append(float(movement_logits[0])) # Vitality assessment vitality_prompts = ["energetic active dog", "lethargic tired dog", "alert playful dog"] inputs = clip_processor(text=vitality_prompts, images=img, return_tensors="pt", padding=True).to(device) with torch.no_grad(): vitality_logits = clip_model(**inputs).logits_per_image.softmax(-1)[0].cpu().numpy() vitality_scores.append(float(vitality_logits[0] + vitality_logits[2])) cap.release() if not health_scores: return None return { "duration_sec": round(total/fps, 1), "mobility_score": float(np.mean(movement_scores)) * 100, "comfort_score": float(np.mean(health_scores)) * 100, "vitality_score": float(np.mean(vitality_scores)) * 100, "frames_analyzed": len(health_scores), "mobility_assessment": "Normal gait pattern" if np.mean(movement_scores) > 0.6 else "Mobility concerns detected", "comfort_assessment": "No obvious discomfort" if np.mean(health_scores) > 0.7 else "Possible discomfort signs", "vitality_assessment": "Good energy level" if np.mean(vitality_scores) > 0.6 else "Low energy observed" } except Exception as e: return None def score_from_response(response, score_mapping): """Extract numeric score from text response""" if not response: return 50 for key, value in score_mapping.items(): if key.lower() in response.lower(): return value return 50 def calculate_hrqol_scores(hrqol_responses): """Convert comprehensive HRQOL responses to 0-100 domain scores""" score_mapping = { "excellent": 100, "very good": 80, "good": 60, "fair": 40, "poor": 20 } domain_scores = {} # Each domain now has one comprehensive question domain_scores["vitality"] = score_from_response( hrqol_responses.get("vitality_comprehensive", ""), score_mapping ) domain_scores["comfort"] = score_from_response( hrqol_responses.get("comfort_comprehensive", ""), score_mapping ) domain_scores["emotional_wellbeing"] = score_from_response( hrqol_responses.get("emotional_comprehensive", ""), score_mapping ) domain_scores["alertness"] = score_from_response( hrqol_responses.get("alertness_comprehensive", ""), score_mapping ) return domain_scores def get_score_color(score): """Return background and text color based on score for better visibility""" if score >= 80: return {"bg": "#4CAF50", "text": "#FFFFFF"} # Green background, white text elif score >= 60: return {"bg": "#FFC107", "text": "#000000"} # Yellow background, black text elif score >= 40: return {"bg": "#FF9800", "text": "#FFFFFF"} # Orange background, white text else: return {"bg": "#F44336", "text": "#FFFFFF"} # Red background, white text def get_healthspan_grade(score): if score >= 85: return "Excellent (A+)" elif score >= 75: return "Very Good (A)" elif score >= 65: return "Good (B)" elif score >= 55: return "Fair (C)" elif score >= 45: return "Poor (D)" else: return "Critical (F)" def show_loading(): """Display loading animation""" return """

🔬 Analyzing Your Dog's Health...

Please wait while we process the image/video and questionnaire data using enhanced AI models.

""" def comprehensive_healthspan_analysis(input_type, image_input, video_input, breed, age, *hrqol_responses): """Enhanced comprehensive analysis with improved biological age prediction""" # Show loading first yield show_loading() # Simulate processing time for enhanced computations time.sleep(3) # Determine which input to use based on dropdown selection if input_type == "Image Analysis": selected_media = image_input media_type = "image" video_path = None elif input_type == "Video Analysis": selected_media = video_input media_type = "video" video_path = video_input else: yield "❌ *Error*: Please select an input type." return if selected_media is None: yield f"❌ *Error*: Please provide a {media_type} for analysis." return # Check if questionnaire is completed if not hrqol_responses or all(not r for r in hrqol_responses): yield "❌ *Error*: Please complete the HRQOL questionnaire before analysis." return # Build HRQOL responses dictionary - Updated for shortened questionnaire response_keys = [] for domain_key, domain_data in HRQOL_QUESTIONNAIRE.items(): for question in domain_data["questions"]: response_keys.append(question["id"]) hrqol_dict = {key: hrqol_responses[i] if i < len(hrqol_responses) else "" for i, key in enumerate(response_keys)} # Calculate HRQOL scores hrqol_scores = calculate_hrqol_scores(hrqol_dict) # Initialize analysis variables video_features = {} breed_info = None enhanced_age_info = None health_aspects = {} # Perform analysis based on media type if media_type == "image": try: detected_breed, breed_conf, health_aspects = classify_breed_and_health(selected_media, breed) # ENHANCED biological age prediction with improved accuracy enhanced_age_info = predict_biological_age_enhanced( selected_media, None, detected_breed, hrqol_scores, age ) breed_info = { "breed": detected_breed, "confidence": breed_conf, "bio_age": enhanced_age_info['biological_age'], "uncertainty": enhanced_age_info['uncertainty'], "high_uncertainty": enhanced_age_info['high_uncertainty'], "vision_quality": enhanced_age_info['vision_quality'], "confidence_factors": enhanced_age_info['confidence_factors'] } except Exception as e: logger.error(f"Image analysis error: {e}") elif media_type == "video": # For video, analyze both movement and extract frame for breed analysis video_features = analyze_video_gait(selected_media) or {} # Try to extract a frame from video for breed analysis try: cap = cv2.VideoCapture(selected_media) ret, frame = cap.read() if ret: img = Image.fromarray(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)) detected_breed, breed_conf, health_aspects = classify_breed_and_health(img, breed) # ENHANCED biological age prediction with video enhanced_age_info = predict_biological_age_enhanced( img, selected_media, detected_breed, hrqol_scores, age ) breed_info = { "breed": detected_breed, "confidence": breed_conf, "bio_age": enhanced_age_info['biological_age'], "uncertainty": enhanced_age_info['uncertainty'], "high_uncertainty": enhanced_age_info['high_uncertainty'], "vision_quality": enhanced_age_info['vision_quality'], "confidence_factors": enhanced_age_info['confidence_factors'] } cap.release() except Exception as e: logger.error(f"Video analysis error: {e}") # Calculate Composite Healthspan Score (enhanced) video_weight = 0.3 if video_features else 0.0 hrqol_weight = 0.7 if video_features else 1.0 if video_features: video_score = ( video_features.get("mobility_score", 70) * 0.4 + video_features.get("comfort_score", 70) * 0.3 + video_features.get("vitality_score", 70) * 0.3 ) else: video_score = 0 hrqol_composite = ( hrqol_scores["vitality"] * 0.25 + hrqol_scores["comfort"] * 0.25 + hrqol_scores["emotional_wellbeing"] * 0.25 + hrqol_scores["alertness"] * 0.25 ) final_healthspan_score = (video_score * video_weight) + (hrqol_composite * hrqol_weight) final_healthspan_score = min(100, max(0, final_healthspan_score)) # Generate comprehensive report with enhanced features input_type_icon = "📸" if media_type == "image" else "🎥" report_html = f"""

{input_type_icon} Enhanced Multi-Modal Health Assessment

Analysis Type: {input_type} | Advanced Biological Age Prediction
{final_healthspan_score:.1f}/100
{get_healthspan_grade(final_healthspan_score)}
""" # Add domain score cards with improved contrast for domain, score in [("vitality", "🔋 Vitality"), ("comfort", "😌 Comfort"), ("emotional_wellbeing", "😊 Emotional"), ("alertness", "🧠 Alertness")]: colors = get_score_color(hrqol_scores[domain]) report_html += f"""

{score.split()[1]}

{hrqol_scores[domain]:.0f}
{hrqol_scores[domain]:.1f}/100
""" report_html += "
" # Enhanced Visual Analysis section with improved accuracy if breed_info: uncertainty_info = "" if breed_info.get('high_uncertainty', False): uncertainty_info = f"""

⚠ High Uncertainty: Age prediction uncertainty is ±{breed_info.get('uncertainty', 0):.1f} years. Consider veterinary consultation.

""" pace_info = "" if age and age > 0: pace = breed_info["bio_age"] / age pace_status = "Accelerated" if pace > 1.2 else "Normal" if pace > 0.8 else "Slow" pace_color = "#FF5722" if pace > 1.2 else "#4CAF50" if pace < 0.8 else "#FF9800" pace_info = f"""

Aging Pace: {pace:.2f}× ({pace_status})

""" vision_quality_info = f"""

Image Quality: {breed_info.get('vision_quality', 0.5)*100:.0f}%

""" # Confidence factors breakdown confidence_factors = breed_info.get('confidence_factors', {}) visual_health = confidence_factors.get('visual_health', {}) hrqol_factor = confidence_factors.get('hrqol_factor', 0) gait_factor = confidence_factors.get('gait_factor', 0) total_health_impact = confidence_factors.get('total_health_impact', 0) prediction_confidence = confidence_factors.get('prediction_confidence', 0.5) factors_info = f"""

Advanced Analysis Factors:

• Visual Health Factor: {visual_health.get('age_factor', 0):.3f}

• HRQOL Adjustment: {hrqol_factor:.3f}

• Gait Factor: {gait_factor:.3f}

• Total Health Impact: {total_health_impact:.3f}

• Prediction Confidence: {prediction_confidence:.1%}

""" report_html += f"""

{input_type_icon} Advanced Visual Analysis

Detected Breed: {breed_info['breed']} ({breed_info['confidence']:.1%} confidence)

Enhanced Biological Age: {breed_info['bio_age']} ± {breed_info.get('uncertainty', 0):.1f} years

Chronological Age: {age or 'Not provided'} years

{vision_quality_info} {pace_info} {factors_info} {uncertainty_info}
""" # Enhanced video analysis if video_features: report_html += f"""

🎥 Advanced Gait & Movement Analysis

Duration: {video_features['duration_sec']} seconds

Mobility Assessment: {video_features['mobility_assessment']}

Comfort Assessment: {video_features['comfort_assessment']}

Vitality Assessment: {video_features['vitality_assessment']}

Enhanced Analysis: {video_features['frames_analyzed']} frames with age-specific movement analysis

""" # Physical Health Assessment with improved visibility if health_aspects and media_type == "image": report_html += f"""

📸 Physical Health Assessment

""" for aspect, data in health_aspects.items(): is_healthy = any(word in data["assessment"].lower() for word in ["healthy", "bright", "clean", "ideal"]) status_icon = "✅" if is_healthy else "⚠" status_color = "#2E7D32" if is_healthy else "#F57C00" bg_color = "#E8F5E8" if is_healthy else "#FFF3E0" report_html += f"""

{status_icon} {aspect}: {data['assessment']} ({data['confidence']:.1%} confidence)

""" report_html += "
" # Enhanced recommendations based on advanced analysis recommendations = [] if hrqol_scores["vitality"] < 60: recommendations.append("🔋 *Vitality Enhancement*: Implement graduated exercise program with monitoring") if hrqol_scores["comfort"] < 70: recommendations.append("😌 *Comfort Support*: Consider pain management and mobility aids") if hrqol_scores["emotional_wellbeing"] < 65: recommendations.append("😊 *Emotional Care*: Increase environmental enrichment and social interaction") if hrqol_scores["alertness"] < 70: recommendations.append("🧠 *Cognitive Support*: Introduce cognitive enhancement activities") if breed_info and breed_info.get('high_uncertainty', False): recommendations.append("🏥 *Veterinary Consultation*: High prediction uncertainty suggests professional evaluation needed") if breed_info and age: pace = breed_info["bio_age"] / age if pace > 1.3: recommendations.append("⚡ *Accelerated Aging*: Consider comprehensive health screening and lifestyle modifications") if recommendations: report_html += f"""

🎯 Enhanced AI Recommendations

{''.join([f'

{rec}

' for rec in recommendations])}
""" # Enhanced disclaimer with model information report_html += """

⚠ Important Disclaimer: This analysis uses advanced AI models with multi-factor biological age prediction based on visual health indicators, HRQOL correlations, and movement analysis. Results are for educational purposes only. Always consult with a qualified veterinarian for professional medical advice and diagnosis.

Advanced Features: Multi-factor age prediction, breed-specific aging rates, enhanced uncertainty quantification, comprehensive health analysis

""" yield report_html def update_media_input(input_type): """Update the visibility of media inputs based on dropdown selection""" if input_type == "Image Analysis": return gr.update(visible=True), gr.update(visible=False) else: # Video Analysis return gr.update(visible=False), gr.update(visible=True) custom_css = """ /* Enhanced gradient background - Orangish fade theme */ .gradio-container { background: linear-gradient(135deg, #ff8a50 0%, #ff6b35 25%, #ff4500 50%, #ff8c00 75%, #ffa500 100%); min-height: 100vh; font-family: 'Inter', -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif; } /* Card styling - Light orange fade background */ .input-card { background: linear-gradient(135deg, #fff4e6 0%, #ffe4cc 100%); border-radius: 12px; padding: 28px; box-shadow: 0 4px 20px rgba(255, 140, 0, 0.15); margin: 12px; border: 1px solid #ffb366; color: #1a202c; } /* Questionnaire grid container - Orange fade design */ .questionnaire-grid { background: linear-gradient(135deg, #fff1e6 0%, #ffe6cc 50%, #ffdbcc 100%); border-radius: 12px; padding: 32px; box-shadow: 0 4px 20px rgba(255, 140, 0, 0.18); margin: 12px; border: 1px solid #ffb366; color: #1a202c; } /* Header styling - Bold orange fade gradient */ .main-header { background: linear-gradient(135deg, #ff6347 0%, #ff7f50 25%, #ff8c00 50%, #ffa500 75%, #ffb347 100%); color: #ffffff; text-align: center; padding: 40px; border-radius: 16px; margin-bottom: 32px; box-shadow: 0 8px 32px rgba(255, 140, 0, 0.3); font-family: 'Inter', -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif; } /* Button styling - Orange fade with depth */ .analyze-button { background: linear-gradient(135deg, #ff6b35 0%, #ff8c00 50%, #ffa500 100%); border: none; color: #ffffff; padding: 16px 32px; font-size: 16px; font-weight: 600; border-radius: 12px; cursor: pointer; transition: all 0.3s ease; box-shadow: 0 4px 16px rgba(255, 107, 53, 0.3); font-family: 'Inter', -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif; } .analyze-button:hover { transform: translateY(-2px); box-shadow: 0 8px 28px rgba(255, 107, 53, 0.4); background: linear-gradient(135deg, #ff5722 0%, #ff7043 50%, #ff8a65 100%); } /* Grid styling for questionnaire */ .question-grid { display: grid; grid-template-columns: 2fr 1fr; gap: 24px; align-items: center; padding: 20px 0; border-bottom: 1px solid #ffcc99; margin-bottom: 16px; } .question-grid:last-child { border-bottom: none; margin-bottom: 0; } /* Orange questionnaire text styling - UPDATED TO MATCH THEME */ .question-text { font-size: 16px !important; color: #e65100 !important; line-height: 1.6 !important; margin: 0 !important; padding-right: 20px !important; font-weight: 500 !important; letter-spacing: 0.025em !important; font-family: 'Inter', -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif !important; } .question-text strong { color: #bf360c !important; font-weight: 600 !important; font-size: 16px !important; font-family: 'Inter', -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif !important; } /* Dropdown styling - Orange fade theme */ .gr-dropdown { border-radius: 8px; border: 2px solid #ffb366; background: linear-gradient(135deg, #fff9f5 0%, #fff4e6 100%) !important; transition: all 0.3s ease; font-size: 14px !important; font-weight: 500 !important; color: #2d3748 !important; font-family: 'Inter', -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif !important; } .gr-dropdown:focus { border-color: #ff8c00; box-shadow: 0 0 0 3px rgba(255, 140, 0, 0.15); outline: none; } .gr-dropdown:hover { border-color: #ff9f43; background: linear-gradient(135deg, #fff7f0 0%, #ffede0 100%) !important; } /* Compact spacing */ .question-section { margin-bottom: 24px; } .question-section:last-child { margin-bottom: 0; } /* Professional headers with orange fade - UPDATED */ .questionnaire-grid h2 { font-size: 28px !important; font-weight: 700 !important; color: #d84315 !important; font-family: 'Inter', -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif !important; } .questionnaire-grid p { font-size: 16px !important; color: #ff6f00 !important; font-weight: 400 !important; font-family: 'Inter', -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif !important; } /* Additional professional styling with orange fade */ .gr-textbox, .gr-number { font-family: 'Inter', -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif !important; color: #2d3748 !important; background: linear-gradient(135deg, #fff9f5 0%, #fff4e6 100%) !important; border: 2px solid #ffb366 !important; border-radius: 8px !important; transition: all 0.3s ease !important; } .gr-textbox:focus, .gr-number:focus { border-color: #ff8c00 !important; box-shadow: 0 0 0 3px rgba(255, 140, 0, 0.15) !important; outline: none !important; } .gr-textbox:hover, .gr-number:hover { border-color: #ff9f43 !important; background: linear-gradient(135deg, #fff7f0 0%, #ffede0 100%) !important; } /* Labels styling - UPDATED TO ORANGE */ label { color: #e65100 !important; font-family: 'Inter', -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif !important; font-weight: 500 !important; font-size: 14px !important; margin-bottom: 8px !important; } /* Media input styling with orange fade */ .gr-image, .gr-video { border-radius: 12px !important; border: 2px solid #ffb366 !important; background: linear-gradient(135deg, #fff9f5 0%, #fff4e6 100%) !important; } /* Additional modern touches with orange fade */ .gr-group { background: transparent !important; border: none !important; } .gr-panel { background: linear-gradient(135deg, #fff9f5 0%, #fff4e6 100%) !important; border: 1px solid #ffb366 !important; border-radius: 12px !important; } """ # Updated Gradio Interface with Orange Questionnaire Font with gr.Blocks( title="🐶 Enhanced AI Dog Health Analyzer", theme=gr.themes.Soft( primary_hue="orange", secondary_hue="amber", neutral_hue="slate", font=["Inter", "system-ui", "sans-serif"] ), css=custom_css ) as demo: # Main Header with Orange Fade gr.HTML("""

🐾 PAWSYears - Every Dog has 20 Years Potential

Your Companion's Next-Gen Health Intelligence Platform

""") with gr.Row(): # Left Column - Enhanced Media Input with gr.Column(scale=1): gr.HTML("""

📸 Media Input Selection

""") # Enhanced dropdown with better styling input_type_dropdown = gr.Dropdown( choices=["Image Analysis", "Video Analysis"], label="🔍 Select Analysis Type", value="Image Analysis", interactive=True, elem_classes=["gr-dropdown"] ) # Media input components with enhanced labels image_input = gr.Image( type="pil", label="📷 Upload Dog Photo or Use Webcam", visible=True, sources=["upload", "webcam"], height=320 ) video_input = gr.Video( label="🎥 Upload Video (10-30 seconds) or Record with Webcam", visible=False, sources=["upload", "webcam"], height=320 ) # Update visibility based on dropdown selection input_type_dropdown.change( fn=update_media_input, inputs=[input_type_dropdown], outputs=[image_input, video_input] ) breed_input = gr.Dropdown( STANFORD_BREEDS, label="🐕 Dog Breed (Auto-detected if not specified)", value=None, allow_custom_value=True, elem_classes=["gr-dropdown"] ) age_input = gr.Number( label="📅 Chronological Age (years)", precision=1, value=None, minimum=0, maximum=25 ) # Right Column - Orange Font HRQOL Questionnaire with gr.Column(scale=1): gr.HTML("""

📋 Health Assessment

Complete all questions for comprehensive healthspan analysis

""") hrqol_inputs = [] # Create compact grid layout with orange text with gr.Group(elem_classes=["question-section"]): for domain_key, domain_data in HRQOL_QUESTIONNAIRE.items(): for i, question in enumerate(domain_data["questions"]): with gr.Row(): with gr.Column(scale=2): gr.HTML(f"""
Q{len(hrqol_inputs)+1}: {question["text"]}
""") with gr.Column(scale=1): dropdown = gr.Dropdown( choices=question["options"], label="", value=None, interactive=True, show_label=False, elem_classes=["gr-dropdown"] ) hrqol_inputs.append(dropdown) gr.HTML("
") # Close questionnaire-grid # Enhanced Analysis Button gr.HTML("""
""") analyze_button = gr.Button( "🔬 Run Advanced AI Health Analysis", variant="primary", size="lg", elem_classes=["analyze-button"] ) gr.HTML("
") # Enhanced Results Section output_report = gr.HTML() # Connect analysis function with loading analyze_button.click( fn=comprehensive_healthspan_analysis, inputs=[input_type_dropdown, image_input, video_input, breed_input, age_input] + hrqol_inputs, outputs=output_report ) # Launch the interface if __name__ == "__main__": demo.launch()