0
$\begingroup$

I'm trying to classify my textile design patterns
(let's just think of it as medieval painting)

what I understand of "multilabel classification" is like this:
it outputs multiple possible result out of all those classes (let's say classes are of some artists, style and technique) so one example could be

  • possible classes: leonardo, artist1, artist2, baroque, renaissance, whatever, oil, dessin, watercolor
  • prediction of img1: leonardo da vinci, artist1, renaissance, oilpainting, wtaercolor

but what I want to do is more like:
- possible classes:
artist style technique
leonardo baroque oil
artist1 renaissance dessin
artist2 whatever watercolor
  • prediction: {artist: ['leonardo', 'artist1'], (maybe they drew it together)
    style: ['renaissance'],
    technique: ['oil', 'watercolor']}

classes are more strictly categorized and but also there could be multiple results from one category of class. I'm not even sure what should it be called and having hard time to find articels for it. Can someone please suggest?

$\endgroup$
1
  • $\begingroup$ So that , You need the resources for multilabel classification. Right? $\endgroup$ – AIFahim Feb 27 at 4:57
0
$\begingroup$

It is not a multiclass problem.

It is a multilabel problem. Since, you have the clusters of classes you want to get. You just let the network predict multiple classes and segregate them afterwards. In this case, you will have single classification head.

Other way to do it, is to separately derive multiple classes of article, technique and style. In this case, you will have three classification heads.

$\endgroup$
2
  • $\begingroup$ Thanks I guess then it measn I will train three model sharing same feature extractor, am i in the right path? $\endgroup$ – Minkoo Kim Feb 27 at 6:14
  • $\begingroup$ Yes, it will be the same backbone but you will have three classification heads. Start trying out the various pretrained models in TF or PyTorch like ResNet, EfficientNet etc. $\endgroup$ – Abhishek Verma Feb 27 at 11:45

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.