I’m not sure if NNs already do this, but perhaps using augmentation on the runtime input might help? Similar to how humans can look at things in different lights or at different angles if needed.
I’m not sure if NNs already do this, but perhaps using augmentation on the runtime input might help? Similar to how humans can look at things in different lights or at different angles if needed.