🐼12.17 Panda Base AI: Joining a Research Group to Improve AI detection for Pandas

This month, I am excited to continue my photo annotation experience at the Cornell Lab of Ornithology in a new direction—analyzing panda behavior using AI. Thanks to Dr. Que Pinjia from my ornithology lab and Dr. Luoli and Libi from the panda veterinary hospital for recommending me to this group, I have joined a research group focused on improving real-time health monitoring for pandas. The goal is to enhance instantaneous response when AI detects potential health issues in pandas (from detecting their behaviors), ultimately ensuring better care and conservation outcomes.

The AI detection system relies on deep learning models, including Faster R-CNN and modified ResNet networks, trained on thousands of images to recognize panda behaviors such as walking, sitting, eating, and climbing. By identifying subtle facial motions with manual annotations—like whether a panda’s eyes or mouth are open—the system can flag abnormalities that may indicate stress or illness. This automated method fills the gaps left by traditional human observation, allowing for 24-hour monitoring with high accuracy. I am really happy to be part of this research, as it connects my previous annotation work with my passion for conservation.