6D Object Pose Regression via Supervised Learning on Point Clouds

Abstract

This paper addresses the task of estimating the 6 degrees of freedom pose of a known 3D object from depth information represented by a point cloud. Deep features learned by convolutional neural networks from color information have been the dominant features to be used for inferring object poses, while depth information receives much less attention. However, depth information contains rich geometric information of the object shape, which is important for inferring the object pose. We use depth information represented by point clouds as the input to both deep networks and geometry-based pose refinement and use separate networks for rotation and translation regression. We argue that the axis-angle representation is a suitable rotation representation for deep learning, and use a geodesic loss function for rotation regression. Ablation studies show that these design choices outperform alternatives such as the quaternion representation and L2 loss, or regressing translation and rotation with the same network. Our simple yet effective approach clearly outperforms state-of-the-art methods on the YCB-video dataset.

Self-Supervised Learning of Non-Rigid Residual Flow and Ego-Motion

“…Evaluation Metrics. We adopt the scene flow evaluation metrics from [7] and the camera pose metrics from [23,5]. We use the following evaluation metrics:…”

Canonical Voting: Towards Robust Oriented Bounding Box Detection in 3D Scenes

“…At inference time, rotations with the highest confidence are used as the output. Both Mahendran et al [12] and Gao et al [4] propose to directly regress rotation from 2D images and 3D point clouds, respectively. offset regression [15]; then a canonical voting algorithm is proposed to generate a vote map on 3D grids; finally, a LCC back projection checking module is leveraged to progressively eliminate false positives and generate bounding boxes from the vote map.…”