hhkda.blogg.se

Specifically, our method consists of threeĬomponents: an identical mapping branch (IMB) for non-shadow regions

Separately and leverage the identical mapping results to guide the shadow To address this problem, we propose to handle these two tasks Such a key issue is not identified by existing deep learning-based shadow Towards only one task instead of both during the training process. Shared weights for these two tasks could lead to the model being optimized We find that these two tasks exhibit poor compatibility, and using I.e., data restoration for shadow regions and identical mapping for non-shadow Pairs, which are desired to complete two distinct tasks via shared weights, Methods train deep neural networks on collected shadow & shadow-free image While leaving non-shadow regions unchanged. Download a PDF of the paper titled Learning Restoration is Not Enough: Transfering Identical Mapping for Single-Image Shadow Removal, by Xiaoguang Li and 5 other authors Download PDF Abstract: Shadow removal is to restore shadow regions to their shadow-free counterparts