There are two points of confusion that I hope you can answer:
1. I have noticed the baseline PSNR has been changed (29.10dB, 29.13dB, 29.00dB). Why is it keep changing and will it be changed again? Because it may affect our subsequent strategy.
2. What are the specific ranking criteria for the test phase? For example, which aspect (runtime, parameters, FLOPs, activations, depths) is the most important one. Will the PSNR be considered at test phase once the method exceeds the performance baseline on validation set? For example, which model (A/B) below will come first?
Model | PSNR on test set | Params | Speed
A | 29.13dB | 4M | 0.1s
B | 29.15dB | 6M | 0.3s
I have the same query, even though I achieved highest PSNR and SSIM but still somehow, I am 28th ranked so kindly let us know what is the ranking criteria here?
Posted by: Sunder_ali @ March 8, 2022, 6:28 a.m.We have the same confusion.
Posted by: AweWoo @ March 28, 2022, 1:13 p.m.