Can I use only one query image per class in Few-shot Learning?

Question

I am studying Few shot Learning. In general, I know that 7 to 15 Query images are allocated for each class.

However, The dataset I use for training has a minimum of 2 and a maximum of 200 pieces per class.

So I have only used one query image.

I have two questions.

1. Is it okay to use only one query image like this?

2. And is it meaningful to see good performance when using fewer queries?

Answer 1

1. Yes it's ok if you use only one query image as your query set.
2. It makes sense if you get better performance using more query instances. In my experience, for a given dataset the accuracy won't improve after a certain threshold if you increase the query set instances.