Spatial Inference Machines

  • Roman Shapovalov ,
  • Dmitry Vetrov ,
  • Pushmeet Kohli

2013 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) |

Published by IEEE

Publication

This paper addresses the problem of semantic segmentation of 3D point clouds. We extend the inference machines framework of Ross et al. by adding spatial factors that model mid-range and long-range dependencies inherent in the data. The new model is able to account for semantic spatial context. During training, our method automatically isolates and retains factors modelling spatial dependencies between variables that are relevant for achieving higher prediction accuracy. We evaluate the proposed method by using it to predict 17-category semantic segmentations on sets of stitched Kinect scans. Experimental results show that the spatial dependencies learned by our method significantly improve the accuracy of segmentation. They also show that our method outperforms the existing segmentation technique of Koppula et al.