4xNomosUniDAT2_box
·
44 commits
to main
since this release
Name: 4xNomosUniDAT2_box
Author: Philip Hofmann
Release Date: 10.09.2023
License: CC BY 4.0
Network: DAT
Scale: 4
Purpose: 4x general purpose upscaler for non-degraded content
Iterations: 110000
epoch: 7
batch_size: 4
HR_size: 128
Dataset: nomos_uni
Number of train images: 2989
OTF Training: No
Pretrained_Model_G: DAT_2_x4
Description: 4x general purpose upscaler for non-degraded content. Has been trained as paired dataset on box downsamples as a DAT2 model, AdamW with pixel, perceptual, gan, color, ldl and ff loss. (am currently training a version of this that could handle jpg compression with multiple downsample algorithms and down_up)
