Euler Angles Based Loss Function for Camera Localization with Deep Learning

This paper concentrates on loss function design for camera localization with deep learning. We directly explore the three Euler angles as the orientation representation in the camera pose regressor. Complicated parameter selection is avoided, against the existing PoseNet and its extended works. Experiments are conducted by using the 7 Scenes datasets and the King's College dataset. Results demonstrate that the proposed approach has better performance in orientation accuracy.