-
-
Detected Animal — Dog AI-based animal detection with 89.69% confidence.
-
Vehicle Proximity — Demo Scene Dog detected near a potential vehicle danger zone.
Inspiration
Animals can face dangerous situations such as roads, nearby vehicles, unsafe areas, and unusual movement patterns. Human monitoring of camera footage can be difficult, especially when potential risks need to be noticed quickly.
We wanted to build an AI-based early-warning system that could identify potential animal welfare risks and clearly explain why an alert was generated.
What it does
PAWSENSE-X — AI Animal Welfare & Danger Intelligence is an explainable AI prototype that combines animal detection, tracking, behavior analysis, danger-zone detection, and environmental risk factors.
The system generates an explainable risk score and highlights potential welfare concerns that may require human attention.
Instead of claiming to diagnose pain, disease, or emotions, PAWSENSE-X focuses on observable risk indicators and recommends human verification.
How we built it
We built PAWSENSE-X using:
- Python
- YOLO-based animal detection with ONNX Runtime
- OpenCV for computer vision
- Animal tracking identification
- Movement and behavior analysis
- Danger-zone detection
- Environmental risk analysis
- Explainable risk scoring
- Event timeline generation
- Streamlit for the interactive dashboard
The main pipeline is:
Animal Detection → Tracking → Behavior Analysis → Danger Zone → Risk Engine → Explainable Alert → Dashboard
Challenges we ran into
One of the main challenges was building a functional computer-vision pipeline while keeping the prototype lightweight and suitable for a hackathon.
Another challenge was making the AI result understandable. Instead of showing only a risk number, we designed the system to display the individual factors contributing to the risk score and provide clear explanations.
Accomplishments that we're proud of
We successfully built an end-to-end working prototype that combines multiple AI and computer-vision components into a single monitoring system.
The prototype can:
- Detect an animal
- Identify its tracking ID
- Analyze movement
- Detect danger zones
- Evaluate environmental risk factors
- Generate an explainable risk score
- Display alerts and an event timeline
- Present the results through a live Streamlit dashboard
We are especially proud of turning these separate components into one unified animal-welfare intelligence pipeline.
What we learned
Building PAWSENSE-X helped us understand how computer vision, behavioral analysis, environmental context, and explainable AI can work together in a practical application.
We also learned how to build and deploy an AI prototype, create an interactive dashboard, manage the project with GitHub, and present AI decisions in a more understandable way.
What's next for PAWSENSE-X
Future versions could include:
- Real-time multi-animal tracking
- Personalized behavioral baselines
- More robust behavioral anomaly detection
- Real-world vehicle and environmental context
- Long-term animal welfare analytics
- Larger and more diverse animal datasets
- Improved real-time monitoring
The long-term vision is to make PAWSENSE-X an AI early-warning layer that can help animal shelters, rescue organizations, veterinary environments, and other animal-welfare applications identify situations that may require human attention.
Log in or sign up for Devpost to join the conversation.