AuraPaws

🐶🐕🐱🐈🐹🐰🐇

Inspiration Animal shelters, rescue networks, and veterinary clinics monitor dozens of animals concurrently, but non-verbal animals often hide distress, pain, and psychological decline until it manifests as overt aggression or severe illness. Silent suffering leads to reduced adoption rates, increased operational costs, and higher euthanasia rates. AuraPaws was created to bridge this communication gap by translating ambient bio-acoustics and subtle physical postures into actionable biometric intelligence.

What it does 💡🔦🔎

AuraPaws is a real-time multimodal monitoring platform that converts ambient audio and camera feeds into a dynamic Predictive Welfare Index (PWI) score (0–100):

Acoustic Stress Profiling: Analyzes vocalization pitch modulation, spectral centroid shifts, and stress-panting cadences to distinguish high-arousal play from acute pain or separation anxiety.

Computer Vision Posture Analysis: Tracks subtle physical indicators, including tail-tuck angles, ear deflection, and repetitive pacing loops.

Veterinary Command Center: Streams live telemetry to an interactive web dashboard, triggering instant alerts when an animal's PWI drops below safe thresholds.

How we built it ⚙️⛏️📐📏🔦🔎

Frontend: Built with Next.js 14 (App Router), TypeScript, and Tailwind CSS.

AI & Processing Engine: Developed a Python FastAPI backend leveraging OpenCV for keypoint tracking and Librosa for audio feature extraction (MFCCs and spectral entropy).

Database & Messaging Layer: Configured Supabase PostgreSQL for structured data storage, utilizing Row-Level Security (RLS) and real-time WebSockets to broadcast critical stress updates to the UI without page refreshes.

Challenges we ran into 😥😟😧😩

Multimodal Data Synchronization: Aligning high-frequency audio buffers with video frame rates required tuning asynchronous workers in FastAPI.

Acoustic Noise Filtering: Differentiating background shelter noise (appliances, human footsteps) from actual animal distress required custom spectral gating.

Real-Time Latency: Optimizing computer vision inference to run efficiently alongside acoustic analysis without introducing dashboard delays.

Accomplishments that we're proud of 😄💹📈💯

Predictive Welfare Index Formula: Successfully engineered a weighted scoring algorithm ($PWI = 100 - [(0.55 \times S_{\text{audio}}) + (0.45 \times S_{\text{vision}})]$) that accurately categorizes animal wellness in real time.Zero-Latency Alerting: Achieving sub-second alert delivery via WebSockets whenever an animal enters critical distress ($PWI < 50$). Clean Hardware Footprint: Designing an enterprise-grade AI system that operates using standard webcams and basic ambient microphones.

What we learned 📚🎒🧠🤓

Biometric Nuances: Subtle posture changes (like ear angles) combined with pitch variance provide a far more reliable indicator of animal stress than relying on vision or audio alone.

Edge Optimization: Streamlining AI inference pipelines is crucial for ensuring accessibility in resource-constrained shelter environments.

What's next for AuraPaws 🐾🐾🐾💞🐕

B2B Shelter Pilot: Deploying pilot setups in local shelters to build baseline behavioral datasets across diverse canine and feline breeds.

Telehealth Integration: Expanding API endpoints for integration into commercial veterinary software and wearable smart collars.

Publication & Open Science: Documenting the system's architecture for publication in the Binnovative Innovation Book series to support global animal welfare research.