CODEWITHSUNDEEP
ibm-watsonvisual-recognition
Richie Close
Richie Close

Reputation: 77

Can you train a custom classifier?

There is some documentation on the IBM site related to visual recognition which is confusing me https://www.ibm.com/watson/developercloud/doc/visual-recognition/customizing.shtml

It states: Because custom classifiers are trained with only the images provided in a single call, you cannot update an existing classifier, you can only create new classifiers or delete existing classifiers.

But in the next section: You can update an existing classifier by adding new classes, or by adding new images to existing classes. To update the existing classifier, use several compressed (.zip) files, including files containing positive or negative images (.jpg, or .png). You must supply at least one compressed file, with additional positive or negative examples.

So is it possible to update a custom made classifier? I'm having issues at the moment trying to do this, but it is possibly due to me being on the trial version?

Upvotes: 0

Views: 280

Answers (3)

Rohit
Rohit

Reputation: 187

Yes we can retrain it but in standard version only not in free version.

Upvotes: 1

Matt Hill
Matt Hill

Reputation: 1106

Yes, you can "retrain" or update a custom classifier by providing additional training data. The conflict in the documentation will be corrected soon, thank you for pointing it out. The first quote was left in by mistake from the first version of the API. The second section is the correct description.

Upvotes: 2

chughts
chughts

Reputation: 4735

I think that is most likely to be a historic documentation problem. When it was first released it was not possible to update classifiers, but this was changed on a subsequent release.

Upvotes: 2

Related Questions