Classifier Declined for Initial Training: Unclear Reasons

[Gabe Frank Calistro]

Could someone let me know how to resolve this issue under the Jobs for triggering a new classifier.  

Classifier 54772 [Source: ####] was declined training, because there weren't enough annotations of at least 2 different labels.

I have 25 images now with 1000 points each across 30 labels and even after the most recent classifier initiation the failure appeared. Does every label need a certain amount of annotations?

Does it need to be at least 8 points of data because the model uses the 1/8th for validation/ training?

I appreciate any help!

[Stephen Chan]

Hello,

Yes, it has to do with the 1/8th for validation, and in this case you also unluckily ran into a quirk of how it works. What it specifically does is take every image whose ID number is divisible by 8. I took a look at this particular source (my admin tools allow me to look it up using the classifier number you posted), and I believe none of the 26 confirmed images happen to have an ID divisible by 8. So, the validation set ended up totally empty.

Try picking one more image to annotate and make sure its image-ID is divisible by 8, and once you finish that, I think that should get it running.