search

GaoshengLiu / LF-DGNet

[TBC 2023] Disparity-Guided Light Field Image Super-Resolution via Feature Modulation and Recalibration
6 stars 1 forks source link

A little doubt #1

Open Hupangzi01 opened 8 months ago

Hupangzi01 commented 8 months ago

Good Job!But I have a little doubt. When training disparity maps in an unsupervised manner, what is the supervision and what is the unsupervised training strategy?Thanks Sincerely!

GaoshengLiu commented 6 months ago

Hi, the depth labels for several real-world datasets are not available. Therefore, we adopt an unsupervised manner to obtain the depth map. The unsupervised training strategy is illustrated in Sec.III-B of the main paper. And the training details are provided in Sec. IV-A.