Not really. There's sound theoretical basis for that approach in astrophotography. The sort of bias and or noise introduced by systemically poor labeling is much less reliably averaged out.