Abstract:

Finding and helping people after a natural disaster is difficult since they need to be located quickly in order to receive aid. For search operations, a rescue team is assembled, and they have to hurry to the impacted region in order to identify and assist an injured or missing person. It takes time to conduct a search by assembling a team and heading to the site, which could result in the injured person dying. To solve this issue unmanned drones are used to take pictures of the affected area. Rescuers could discover a lost or wounded person rapidly through these photographs that have been processed by certain machine learning algorithms that recognize the individual and their location. Small proportions of humans are evident in most of the images because these images are usually taken from high elevations. We present a model based on the RetinaNet and ResNet architecture for the detection of humans in aerial images

Description:

Search and Rescue, Unmanned drones, object detection, RetinaNet, Deep Learning, ResNet