Current artificial intelligence methods for object recognition work exclusively with undistorted images. However, 360° images lead to significant image distortions. These distortions hinder accurate object recognition.
The overarching goal of this project is to develop an algorithm for efficient object recognition in 360° data, enabling the direct detection of objects within the distorted images.
The developed methods represent initial solutions demonstrating how existing AI solutions can be optimally adapted to 360° content with minimal effort.