search

yunxiaoshi / Neural-IMage-Assessment

A PyTorch Implementation of Neural IMage Assessment
Other
526 stars 95 forks source link

Creating the dataset partitions #8

Closed prashanthvs closed 5 years ago

prashanthvs commented 5 years ago

Can you please share the CSV files that you used to generate the data partitions for train, val and test.

yunxiaoshi commented 5 years ago

Hi,

I don't seem to have the csv files locally anymore, but creating them is fairly easy, say the original info files look like this (after some simple preprcocessing)

2 2 3 4 12 21 22 12 4 6
...

each number stands for the number of people voted for that score bucket, than you just divide them by their sum to form a empirical distribution,

0.02273 0.02273 0.03409 0.04545 0.13636 0.23863 0.13636 0.04545 0.06818
...

then you can go ahead to create your train/test split

KevinTech0916 commented 5 years ago

img_path csv_file how can I determine these, according to AVA.txt? I am not very familiar with data processing. Need your help Thank you ! your code is awesome !

yunxiaoshi commented 5 years ago

It is now provided in README