3つの要点 ✔️ 低解像から高解像へ変換 ✔️ 時間的に一貫したフレームを生成するために双方向損失関数を採用 ✔️ 超解像度へ変換するTecoGANを提案 Learning Temporal Coherence via Self-Supervision for GAN-based Video Generation Temporally Coherent GANs for Video Super-Resolution (TecoGAN) written by Mengyu Chu, You Xie, Jonas Mayer, Laura Leal-Taixé, Nils Thuerey (Submitted on 23 Nov 2018 (v1), last revised 21 May 2020 (this version, v4)) Comments: Published by arXi