Homography-Based Loss Function for Camera Pose Regression - Archive ouverte HAL Access content directly
Journal Articles IEEE Robotics and Automation Letters Year : 2022

Homography-Based Loss Function for Camera Pose Regression

Abstract

Some recent visual-based relocalization algorithms rely on deep learning methods to perform camera pose regression from image data. This paper focuses on the loss functions that embed the error between two poses to perform deep learning based camera pose regression. Existing loss functions are either difficult-to-tune multi-objective functions or present unstable reprojection errors that rely on ground truth 3D scene points and require a two-step training. To deal with these issues, we introduce a novel loss function which is based on a multiplane homography integration. This new function does not require prior initialization and only depends on physically interpretable hyperparameters. Furthermore, the experiments carried out on well established relocalization datasets show that it minimizes best the mean square reprojection error during training when compared with existing loss functions.
Fichier principal
Vignette du fichier
RA_L_2022_clean.pdf (992.76 Ko) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-03654445 , version 1 (03-05-2022)

Identifiers

Cite

Clémentin Boittiaux, Ricard Marxer, Claire Dune, Aurélien Arnaubec, Vincent Hugel. Homography-Based Loss Function for Camera Pose Regression. IEEE Robotics and Automation Letters, In press, 7 (3), pp.6242-6249. ⟨10.1109/LRA.2022.3168329⟩. ⟨hal-03654445⟩
155 View
147 Download

Altmetric

Share

Gmail Facebook X LinkedIn More