Within my images I have some objects labeled as “Don’t Care”; where I do not want the error to be counted during training. Is that possible to do?
Per your description, there is a “Don’t Care” class in your dataset.
What do you mean by “error”? What’s the detailed about “error”?
If you do not want to train “Don’t Care”, yes, it is possible and easy. Just not set this class in the training spec.