Deep Transformer Network for Monocular Pose Estimation of Shipborne Unmanned Aerial Vehicle

What is it about?

The article presents a deep learning model, the Transformer Neural Network for Multi Object (TNN-MO), designed to estimate the 6D pose of a UAV relative to a ship using monocular images. The model detects keypoints on multiple ship parts and integrates these estimates using Bayesian fusion, showing high accuracy in various lighting conditions. This technology could significantly improve autonomous UAV landing and navigation on ships without relying on GPS. The article also presents details the creation of virtual environments and the generation of synthetic data for training the model.

Featured Image