Feature pyramids and iterative refinement have recently led to great progress in optical flow estimation . However, downsampling in feature pyramids can cause blending of foreground objects with background . We propose anovel Residual Feature Pyramid Module (RFPM) which retains important details inthe feature map without changing the overall design of the overall iterative . RFPM incorporates a residual structure between multiple feature pyramid structures into a downsampled module that corrects the blending of objects across boundaries . It is one of the top-performing methods in KITTI. Results show that ourRFP M visibly reduces flow errors and improves state-of-art performance in the clean pass of Sintel, and is a top-performing method in KITI. It is a special learning approach that can

Author(s) : Libo Long, Jochen Lang

Links : PDF - Abstract

Code :
Coursera

Keywords : feature - residual - flow - pyramid - rfpm -

Leave a Reply

Your email address will not be published. Required fields are marked *