Visual interpretation of Figure 8(b): the high-scale σ_g=36 setting has lower test PSNR than the intermediate σ_g=12 setting even with feature training, consistent with the reported lack of benefit from joint optimization. · CiteArk
Visual interpretation of Figure 8(b): the high-scale σ_g=36 setting has lower test PSNR than the intermediate σ_g=12 setting even with feature training, consistent with the reported lack of benefit from joint optimization.
