Open tkuri opened 3 years ago
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。
https://openaccess.thecvf.com/content_CVPR_2020/html/Song_Channel_Attention_Based_Iterative_Residual_Learning_for_Depth_Map_Super-Resolution_CVPR_2020_paper.html
未確認。
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。
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。