researchmm / ckdn Goto Github PK

Thanks for your excellent work!
I would like to ask whether only the training phase needs LR as a reference, and a single test should only need to provide SR as input.
But when I used predict_score.py, I found that it refers to “timm.data.dataset”， It still needs to deal with data like training or val phase.
What should I do if I want to test my own pictures, or what should be the input form?
I'd appreciate it if you would tell me~

Thanks for your excellent work!
Would you please tell me more about the training details? What exactly are the datasets?
I'd appreciate it if you would tell me~

Hi, as seen in the train.py and ckdn.py files, the test phase also used reference images that were inconsistent with the original paper description. Or did I miss the right?

use which code to generate degraded images?
can you provide the image process codes?

The target should be removed.

In Table 2 of your paper, did you re-train the NR-IQA models on your dataset or use their pre-trained model for evaluation?

In fact, some previous studies have proposed the use of degraded images for quality evaluation.
[1] No-reference image contrast evaluation by generating bi-directional pseudo references
[2] Tspr: Deep network-basedblind image quality assessment using two-side pseudo reference images