Reimplementation on FreiHand

[BenjaminJonghyun]

Hello,
I'm trying to reimplementate your work with A100*8EA.
However, performance was not reached to your one.
I found that some configurations are different compared with your training log.
Could you share more details about training configurations or exact training scripts?

[delaprada]

I also have the same problem. The result can only achieve 6.9 mm for one GPU using the default configuration. There are some configuration differences between the log and the code. I am not sure whether the configuration matters.