Award for Reproducing the Paper Results

Open zcy8123878 opened this issue 4 years ago • 3 comments

Hi,

If ANYONE succeeded in reproducing the paper results, you are welcome to send your training code along with your huobi wallet address to me via [email protected]. After checking the results myself, I will send 0.3ETH(≈800$) to your wallet but only for the first provider : )

Thanks ~

Jun 05 '21 06:06 zcy8123878

you'd better train your net on 3ddfa_v1, v2's pipeline is too complicated.

Jun 08 '21 06:06 lyyiangang

I have no difficulty in representing v1.0, however, I made much effort but failed to represent v2.0, some methods in the paper just didn't work well for me.

Jun 08 '21 07:06 zcy8123878

The benefit of meta-joint optimization is limited, more augmentation produces better results. If possible, using externel data (the paper only uses 300W-LP) may promote the generalization ability. @zcy8123878

Jan 23 '22 10:01 cleardusk