search

zwx8981 / LIQE

[CVPR2023] Blind Image Quality Assessment via Vision-Language Correspondence: A Multitask Learning Perspective
MIT License
197 stars 11 forks source link

The pretrained model is trained on which dataset? #6

Closed liuppboy closed 1 year ago

liuppboy commented 1 year ago

Hi, may I ask which dataset do you use for training? Is the score range from 1 to 5? Thanks!

zwx8981 commented 1 year ago

@liuppboy Hi, thanks for your interest in our work.

1, Similar to our previous work UNIQUE (TIP2021), the pre-trained model is jointly trained on a combination of the training sets from six datasets (LIVE, CSIQ, BID, LIVE Challenge, KonIQ-10k, and KADID-10k) under the learning-to-rank framework.

2, Yes, the score ranges from 1 to 5.