I am new to few-shot learning, and I wanted to get a hands-on understanding of it, using Reptile algorithm, applied to my custom dataset.

My custom dataset has 30 categories, with 5 images per category, so this would be a 30 way 5 shot.

Given a new image, I wish to be able to classify it into one of 30 categories. I changed train_shots = 5, classes = 30 in the linked example, and got the training output as

batch 0: train=0.050000 test=0.050000
batch 1: train=0.050000 test=0.050000

Should the custom dataset be used as a validation set, with mini-ImageNet as a training dataset, so that the knowledge is transferred? Or can I use only a custom dataset with only $30*5=150$ images for training?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.