search

zwx8981 / UNIQUE

The repository for 'Uncertainty-aware blind image quality assessment in the laboratory and wild' and 'Learning to blindly assess image quality in the laboratory and wild'
Apache License 2.0
127 stars 10 forks source link

About the dataset that is lack of the variance data #12

Closed SuperCrystal closed 3 years ago

SuperCrystal commented 3 years ago

Hello, the model seems work very well on dataset such as tid2013, koniq10k etc. But there is a problem that what if the variance of MOS data is unknown?

zwx8981 commented 3 years ago

@SuperCrystal Hi, thank you for your interest. If the variance of MOS data is unknown, you can just use the binary labels to train the model.

SuperCrystal commented 3 years ago

@SuperCrystal Hi, thank you for your interest. If the variance of MOS data is unknown, you can just use the binary labels to train the model.

Thanks a lot. The binary labels seem work fine to me:)