Winning Solution in NTIRE19 Challenges on Video Restoration and Enhancement (CVPR19 Workshops) - Video Restoration with Enhanced Deformable Convolutional Networks. EDVR has been merged into BasicSR and this repo is a mirror of BasicSR.
Hello! I use my own data as the training set to retrain the network. Because the data is 16bits gray-scale data, when normalizing, I simply divide the original network by 255 and change it to 65535 and set the input channel =1. The training strategy and parameter configuration are basically unchanged. When testing data, for the temporal Attention, I get cor_prob which values are basically greater than 0.998.
How to explain this phenomenon? Does it have anything to do with my training set? Do I need to make any other changes when I change the data set? Or is there a problem with my network training?
Hello! I use my own data as the training set to retrain the network. Because the data is 16bits gray-scale data, when normalizing, I simply divide the original network by 255 and change it to 65535 and set the input channel =1. The training strategy and parameter configuration are basically unchanged. When testing data, for the temporal Attention, I get cor_prob which values are basically greater than 0.998.
How to explain this phenomenon? Does it have anything to do with my training set? Do I need to make any other changes when I change the data set? Or is there a problem with my network training?
thank you!