Adding even a small amount of real images to the training sets should be beneficial. The current idea is to do automated annotation using current models, then do human review.
Adding even a small amount of real images to the training sets should be beneficial.
The current idea is to do automated annotation using current models, then do human review.