tkuri / papers

Summarizing the papers I have read (Japanese)
41 stars 1 forks source link

Channel Attention Based Iterative Residual Learning for Depth Map Super-Resolution #358

Open tkuri opened 3 years ago

tkuri commented 3 years ago

論文概要

Depth画像の超解像。高解像度(HR)から低解像度(LR)への劣化をノイズを含む非線形ダウンサンプリング・インターバルダウンサンプリングの2種類に分類し、それぞれに別個にモデルを学習することが性能向上に寄与すると指摘。またTGV正則化をしてDepthのエッジをはっきりさせる。

bib_20210305 00 https://openaccess.thecvf.com/content_CVPR_2020/html/Song_Channel_Attention_Based_Iterative_Residual_Learning_for_Depth_Map_Super-Resolution_CVPR_2020_paper.html

Code

未確認。