Fusing Additional input/features for Transfer Learning

Thank you so much for getting back!

Additional data is numeric. Features would be 2 to 4 dimensional, definitely not more than that.

So after going through the vision forum, I think I found some other posts that were very similar to what I was trying to do. I am referring to this post specifically. But then I have some other questions from there as well. I have a total of 250 something images over two classes. Classes are named, PD and Non_PD.

So let’s say I have csv file with the header: filename, feature_1, feature_2, feature_3, feature_4. Now the files in filename are all located in PD and Non_PD folders but there are no indications in the csv file which are in which class (I could add it if it would make my life easy).

That being said, for my problem what would additional_data_dimension be as referred to in the other post I just mentioned? 4? How would I read in the data here? Through DataLoader? Since right now my code is getting all the images from the train directory. How would the code look like for that?