ASPP-DF-PVNet: Atrous Spatial Pyramid Pooling and Distance-Filtered PVNet for occlusion resistant 6D object pose estimation

作者：

Highlights：

• An occlusion resistant framework for 6D object pose estimation is proposed.

• The ASPP module is integrated into the network to get more accurate segmentation and vector-field predictions.

• The voting results are improved by taking the voting distance into account to filter out the votes with large deviations.

摘要

•An occlusion resistant framework for 6D object pose estimation is proposed.•The ASPP module is integrated into the network to get more accurate segmentation and vector-field predictions.•The voting results are improved by taking the voting distance into account to filter out the votes with large deviations.

论文关键词：6D object pose estimation,Vector fields,Voting based keypoint localization,Semantic segmentation,ASPP

论文评审过程：Received 1 September 2020, Revised 27 February 2021, Accepted 7 April 2021, Available online 19 April 2021, Version of Record 19 April 2021.

论文官网地址：https://doi.org/10.1016/j.image.2021.116268