Article: Rotation-invariant 3D convolutional neural networks for 6D object pose estimation Journal: International Journal of Computational Science and Engineering (IJCSE) 2025 Vol.28 No.8 pp.1 - 9 Abstract: 6D object pose estimation, crucial for applications such as scene understanding, AR/VR, and robotic grasping, focuses on determining an object's rotation and translation from single-view images. Despite advancements in 3D deep learning, existing methods still struggle with large shape variations, high training demands, and unseen poses. This paper addresses these issues by introducing a rotation-invariant neural network. We propose a rotation-invariant 3D convolutional network that processes a point cloud to predict per-point canonical coordinates, from which the 6D pose, is estimated. The network utilises relative distances and angles within a representative point set. Experiments on a public dataset show that our method outperforms several state-of-the-art baselines, excelling in handling novel poses and severe occlusions. Ablation studies further highlight the importance of the individual components. Inderscience Publishers - linking academia, business and industry through research

Title: Rotation-invariant 3D convolutional neural networks for 6D object pose estimation

Authors: Zhizhong Chen; Zhihang Wang; Xue Hui Xing; Tao Kuai

Addresses: Northwest Institute of Mechanical and Electrical Engineering, No. 5 Biyuan East Road, Weicheng District, Xianyang City, Shaanxi Province, 712000, China ' Northwest Institute of Mechanical and Electrical Engineering, No. 5 Biyuan East Road, Weicheng District, Xianyang City, Shaanxi Province, 712000, China ' Northwest Institute of Mechanical and Electrical Engineering, No. 5 Biyuan East Road, Weicheng District, Xianyang City, Shaanxi Province, 712000, China ' Northwest Institute of Mechanical and Electrical Engineering, No. 5 Biyuan East Road, Weicheng District, Xianyang City, Shaanxi Province, 712000, China

Abstract: 6D object pose estimation, crucial for applications such as scene understanding, AR/VR, and robotic grasping, focuses on determining an object's rotation and translation from single-view images. Despite advancements in 3D deep learning, existing methods still struggle with large shape variations, high training demands, and unseen poses. This paper addresses these issues by introducing a rotation-invariant neural network. We propose a rotation-invariant 3D convolutional network that processes a point cloud to predict per-point canonical coordinates, from which the 6D pose, is estimated. The network utilises relative distances and angles within a representative point set. Experiments on a public dataset show that our method outperforms several state-of-the-art baselines, excelling in handling novel poses and severe occlusions. Ablation studies further highlight the importance of the individual components.

Keywords: rotation-invariant; geometric feature extraction; pose estimation.

DOI: 10.1504/IJCSE.2025.145133

International Journal of Computational Science and Engineering, 2025 Vol.28 No.8, pp.1 - 9

Received: 20 Sep 2024
Accepted: 19 Feb 2025
Published online: 20 Mar 2025 *

Title: Rotation-invariant 3D convolutional neural networks for 6D object pose estimation

Keep up-to-date