I'm thrilled to announce that I've been recognized as a winner in the Moodeng AI Challenge for Track 3: AI for Augmenting Zoo Keepers! My project, MooDong, utilizes multi-task LSTM and vision models to extract pose, mood, hunger, and future-movement predictions for animals, all without requiring manual labels.

The Moodeng AI Challenge, hosted by MIT Media Lab, focuses on innovative AI solutions for real-world problems. My solution addresses the critical need for efficient and non-invasive animal monitoring in zoos, providing keepers with actionable insights into animal well-being.

About MooDong

MooDong is designed to revolutionize how zoo keepers monitor animal health and behavior. By leveraging advanced AI models, it can:

• Accurately detect animal poses and movements.
• Infer mood and emotional states from behavioral patterns.
• Predict hunger levels based on activity and feeding schedules.
• Forecast future movements to anticipate needs or potential issues.

Key Technologies

• Multi-task LSTM: For sequential data analysis and prediction.
• Vision Models: For real-time pose estimation and behavioral analysis.
• Unsupervised Learning: Eliminates the need for extensive manual labeling, making the system highly scalable and adaptable.

Impact on Zoo Keeping

MooDong provides zoo keepers with a powerful tool to enhance animal welfare, optimize care routines, and detect early signs of distress or illness. This leads to healthier animals and more efficient operations.