Multi-Level Fusion Net for hand pose estimation in hand-object interaction

作者：

Highlights：

• Hand pose estimation model for hand interacting with an unknown object.

• A multi-level information fusion framework: “Resolution, Modal and Focus” fusion.

• One RGBD camera setup at free viewpoint.

• Fast inference of 290fps and competitive accuracy on FHAD dataset.

摘要

•Hand pose estimation model for hand interacting with an unknown object.•A multi-level information fusion framework: “Resolution, Modal and Focus” fusion.•One RGBD camera setup at free viewpoint.•Fast inference of 290fps and competitive accuracy on FHAD dataset.

论文关键词：Hand pose estimation,Hand–object interaction,Occlusion,RGBD,Convolutional Neural Networks

论文评审过程：Received 8 August 2019, Revised 8 January 2021, Accepted 7 February 2021, Available online 11 February 2021, Version of Record 24 February 2021.

论文官网地址：https://doi.org/10.1016/j.image.2021.116196