search

weizhou-geek / Tensor-NLFQ

Demo code of "Tensor Oriented No-Reference Light Field Image Quality Assessment", TIP 2020
5 stars 2 forks source link

The problem of NAN feature on Win5-LID dataset. #3

Closed ZhengyuZhang96 closed 3 years ago

ZhengyuZhang96 commented 3 years ago

Hi, author! Thanks for your attractive work and code. Now I am trying to train a model on Win5-LID dataset, but there are 4 images whose features contian "NAN" after feature extraction with "Tensor-NLFQ/main.m". Specifically, these four images are: NN_greek_40, NN_dishes_40, NN_museum_40, NN_rosemary_40.

Since NAN feature will lead to training errors, how to solve this problem? Looking forward to your reply! Thanks!

weizhou-geek commented 3 years ago

Hi, author! Thanks for your attractive work and code. Now I am trying to train a model on Win5-LID dataset, but there are 4 images whose features contian "NAN" after feature extraction with "Tensor-NLFQ/main.m". Specifically, these four images are: NN_greek_40, NN_dishes_40, NN_museum_40, NN_rosemary_40.

Since NAN feature will lead to training errors, how to solve this problem? Looking forward to your reply! Thanks!

Hi, there! Thanks for your interest. Actually, you can set the NAN to zero or random number, and then test the results.

ZhengyuZhang96 commented 3 years ago

Thanks a lot !