Some problems about the results

zhuhancheng / MetaIQA

126 stars 27 forks source link

Some problems about the results #4

Closed Shuweis closed 4 years ago

Shuweis commented 4 years ago

Hi!Congratulations to your work! In the process of reproducing the code, I trained the prior knowledge model on TID2013 and Kadid10k, and fine-tuned on the CLIVE data set, but the SROCC was about 0.48, which did not achieve the results in the paper. Do you get similar results? I didn't make any change but the results were not ideal.

zhuhancheng commented 4 years ago

This issue has been fixed by commenting out 'labels= labels / 100.0'.