Paper Image

Learning object poses from spherical representations

Published on:

19 August 2023

Primary Category:

Computer Vision and Pattern Recognition

Paper Authors:

Jiehong Lin,

Zewei Wei,

Yabin Zhang,

Kui Jia


Key Details

Proposes VI-Net to factorize rotation learning into viewpoint and in-plane branches

Learns viewpoint rotation via binary classification on the sphere

Transforms features to estimate in-plane rotation from canonical view

Introduces spatial spherical convolutions for continuous feature extraction

Achieves state-of-the-art category-level 6D object pose estimation

AI generated summary

Learning object poses from spherical representations

This paper proposes a network called VI-Net that estimates precise 6D object poses by factorizing rotation learning into two branches handling viewpoint and in-plane rotations on spherical representations. Without CAD models, it achieves state-of-the-art performance.

Answers from this paper


No comments yet, be the first to start the conversation...

Sign up to comment on this paper

Sign Up