Das Paper mit dem Titel "Towards Learning Monocular 3D Object Localization Using the Physical Laws of Motion" von Daniel Kienzle, Julian Lorenz, Katja Ludwig und Rainer Lienhart wurde auf der Internation Conference on 3D Vision (3DV) 2024 akzeptiert. Das Paper beschreibt eine neue Methode, um Objekten in 3D zu lokalisieren, ohne dass 3D-Grundwahrheiten benötigt werden. Stattdessen wird das Wissen über physikalische Gesetze genutzt, um diese Aufgabe zu lernen.

 

Weitere Informationen zu diesem Paper sind unter https://kiedani.github.io/3DV2024/ zu finden.

 

We present a novel method for precise 3D object localization in single images from a single calibrated camera using only 2D labels. No expensive 3D labels are needed. Thus, instead of using 3D labels, our model is trained with easy-to-annotate 2D labels along with the physical knowledge of the object's motion. Given this information, the model can infer the latent third dimension, even though it has never seen this information during training. Our method is evaluated on both synthetic and real-world datasets, and we are able to achieve a mean distance error of just 6 cm in our experiments on real data. The results indicate the method's potential as a step towards learning 3D object location estimation, where collecting 3D data for training is not feasible.