I tried using random images for hand signs in the model, but it gives out wrong outputs for simple signs, whereas it gives right answers for the training data. How do I fix this?
Hi kashera,
This appears to be some kind of bug. People at the backend are currently working on this issue.