Spaces:
Running
More charaters?
It would seem that it doesn't recognize some characters. Can you please increase the amount of characters?
Not on those pre-trained model, no.
One thing that somebody willing might play with, though, is to perform feature extraction and train one or more (small) dense layers for different characters and keep a repo of pretrained classifiers. Or perform contrastive training, so the user only needs a bunch of reference images of their target characters and the network outputs a yes/no answer when asked whether if character x is in image y.
Not on those pre-trained model, no.
One thing that somebody willing might play with, though, is to perform feature extraction and train one or more (small) dense layers for different characters and keep a repo of pretrained classifiers. Or perform contrastive training, so the user only needs a bunch of reference images of their target characters and the network outputs a yes/no answer when asked whether if character x is in image y.
That's waht exactly I want to do with this kinda modals, could you plz share some guide or any advice about that ? Thanks!
Not on those pre-trained model, no.
One thing that somebody willing might play with, though, is to perform feature extraction and train one or more (small) dense layers for different characters and keep a repo of pretrained classifiers. Or perform contrastive training, so the user only needs a bunch of reference images of their target characters and the network outputs a yes/no answer when asked whether if character x is in image y.That's waht exactly I want to do with this kinda modals, could you plz share some guide or any advice about that ? Thanks!
I would love to know more about that too! it would be so much easier to create image datasets of new characters.