Ignore tags with less than 5 images when training
Currently, if you want to train your project, all tags need to have at least 5 images otherwise an error will be thrown and your entire project will not get trained.
Please allow an option to ignore tags with less that 5 images when training
This feature is supported, but the documentation needs updating. They expect this will happen in the next 2-4 weeks.
Juho Hanhimäki commented
This issue makes the workflow of building model / tags dynamically from user inputs really cumbersome. Please improve.
Jonathan Zwart commented
Not just ignore tags with < 5 images - but the ability to select an arbitrary subset of tags for Training.
Completely agree with this. We would like to upload our tag taxonomy first, then upload images and train them. When tags are underused (because my image set may be incomplete at the moment) they should be ignored. Right now I have to delete underused tags then it becomes a difference management mess.
Tim Decker commented
Allow us to set a minimum threshold to include in training. Default it to something reasonable to prevent it from distorting the results.