papers AI Learner
The Github is limit! Click to go to the new site.

SilhoNet: An RGB Method for 6D Object Pose Estimation

2019-03-06
Gideon Billings, Matthew Johnson-Roberson

Abstract

Autonomous robot manipulation involves estimating the pose of the object to be manipulated. Methods using RGB-D data have shown great success in solving this problem. However, there are situations where cost constraints or the working environment may limit the use of RGB-D sensors. When limited to monocular camera data only, the problem of object pose estimation is very challenging. In this work, we introduce a novel method called SilhoNet that predicts 6D object pose from monocular images. We use a Convolutional Neural Network (CNN) pipeline that takes in region of interest proposals to simultaneously predict an intermediate silhouette representation for objects with an associated occlusion mask and a 3D translation vector. The 3D orientation is then regressed from the predicted silhouettes. We show that our method achieves better overall performance than the state-of-the art PoseCNN network for 6D pose estimation.

Abstract (translated by Google)
URL

http://arxiv.org/abs/1809.06893

PDF

http://arxiv.org/pdf/1809.06893


Similar Posts

Comments