I'm reproducing DiffSwap as a baseline for a research, but the results is not coherent with the posted smthngs in the manuscript.
Do you know any tip or techniques for getting plausible results?
Even the identity score is around 0.2. compared with SimSwap and InfoSwap which have about 0.5.
I'm reproducing DiffSwap as a baseline for a research, but the results is not coherent with the posted smthngs in the manuscript. Do you know any tip or techniques for getting plausible results? Even the identity score is around 0.2. compared with SimSwap and InfoSwap which have about 0.5.
How would you think about it? :)
All The Best, Jaeseong Lee