weighted_DCNN_IQA icon indicating copy to clipboard operation
weighted_DCNN_IQA copied to clipboard

the cross dataset test problem

Open badaodao opened this issue 5 years ago • 0 comments

Hello,wonderful u!

I want to train the model on the LIVE dataset and test on TID2013. However,they have different label.(dmos for LIVE,mos for TID2013). I noticed that some articles will give that:

"A nonlinear mapping based on a logistic function(always applied to FR measures) is applied for transforming the quality measure into a certain range.We randomly split the TID2013 into 80% and 20%, 80% is used for estimating parameters of the logistic function and 20% for testing the transformed prediction scores." So I find the FR measures transform method in http://sse.tongji.edu.cn/linzhang/IQA/IQA.htm (I don't know whether you could open this link...)

However,when I use the above method, the test results SROCC is so low and so different with the other cross dataset test results. I want to ask if you know the certain method of the transform the DMOS to MOS? or any advice for me? Thank you very much!!!

badaodao avatar Aug 15 '18 08:08 badaodao