search

lidq92 / msmlTMIQA

Quality Assessment for Tone-Mapped HDR Images Using Multi-Scale and Multi-Layer Information (ICMEw 2018)
6 stars 2 forks source link

Demo or Sample examples #1

Open harshithdwivedi opened 3 years ago

harshithdwivedi commented 3 years ago

Hi @lidq92, great work as always! I was wondering if you had a demo or some sample examples showcasing how the model performs across two similar images?

Thanks!

lidq92 commented 3 years ago

Hi @harshithdwivedi, Thank you for your attention. Currently, I am busy with my PhD thesis, and here are the comments about your issue. You can save the PLSR model parameters betaR in line 77 of QualityPrediction.m (for example, save the model on the 1st split experiment), then refer to line 90 for the inference stage. More specifically, assume you have two images A and B and you have saved the trained PLSR model parameters betaR, you can extract their features F in the form of 2xf matrix by calling the function SFAfeature.m, then apply the inference stage of the PLSR model with [ones(2,1) F]*betaR, finally, you will see the predicted quality scores of images A and B.

When I am free, I may upload a demo (as well as simplify unnecessary code and improve the readability of the code). But with these comments, you should be able to write the demo for your situation. A pull request is welcomed.