Detail about training about stdf-pytorch HOT 2 CLOSED

Hi

1. Yes. BTW I'm not the author of this paper. I re-trained QP=37 model, and then open-sourced it.
2. My reproduced model is STDF-R3, not R3L. My reproduced result is 0.778 dB, which is slighly higher than the reported result 0.75 dB.
3. Some reviewers think that all compared approaches should be re-trained over the same dataset. However, to some reviewers, the generalization ability is also an attribute of an approach. Therefore, the author may not need to re-train the compared approach.

from stdf-pytorch.

Okay, thank you very much for your reply.

from stdf-pytorch.