(AD-POLARIS) Polarization Imaging through Turbid Media

This open-access dataset provides polarized images captured through turbid media with annotations of human and mannequin bounding boxes. The dataset allows researchers to develop and evaluate vision algorithms in challenging scenarios where traditional RGB vision is ineffective. The inclusion of both humans and mannequins provides opportunities for differentiating real people from objects mimicking human shapes, crucial for applications like security and search and rescue. The dataset also includes YOLO-like bounding boxes for both people and mannequins.