id	author	title	date	pages	extension	mime	words	sentence	flesch	summary	cache	txt
ispiv-131	Kurinara, Daiki; Blois, Gianluca; Sakaue, Hirotaka; Schiavazzi, Daniele	Investigating Optimal Training and Uncertainty Quantification for CNN-based Optical Flow	2021	2	.pdf	application/pdf	1108	46	40	The first newly tested training approach consisted of multiple sessions where finer image resolution levels are progressively activated at each successive stage (staged training, see Fig. 1(a)). The network progressively identifies flow features in a coarse-to-fine resolution pipeline, starting from level 6 (coarser resolution) to level 2 (finer resolution).	cache/ispiv-131.pdf	txt/ispiv-131.txt