Aerial Imagery Reconnaissance for Collaborative Search & Rescue
We developed a semantic segmentation model using UNET for the detection of humanoid targets in natural disaster situations. Featuring multi-agent collaboration, our simulation delegates heterogeneous tasks to a UAV and ground rover, prompting them to collaboratively locate survivors, evaluate their health, and coordinate with human rescue services. We employed automated semantic labeling to aid with transfer learning for real-world application success, facilitating massive synthetic image data generation in a timely and cost-effective manner.
Interns: Annchi Liu, Dominic Mullen, and Kaelyn Sun