search

taewookim0812 / C-3PO_Motion_Retargeting_Module

C-3PO_Motion_Retargeting_Module
9 stars 5 forks source link

trained_models link error #13

Open sunbin1357 opened 3 years ago

sunbin1357 commented 3 years ago

Where can I download the trained_models

taewookim0812 commented 3 years ago

I'm sorry. The download link was broken. You can download the trained_models in the followin link.

https://drive.google.com/file/d/1Covr-vyPli7j10EtLmKCN2-WjDzY-uLw/view?usp=sharing

sunbin1357 commented 3 years ago

Thank you for your reply.

Have you tried using deep learning without Reinforcement Learning？Is there any literature on that？

twkim @.***> 于2021年3月22日周一 下午4:36写道：

I'm sorry. The download link was broken. You can download the trained_models in the followin link.

https://drive.google.com/file/d/1Covr-vyPli7j10EtLmKCN2-WjDzY-uLw/view?usp=sharing

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/gd-goblin/C-3PO_Motion_Retargeting_Module/issues/13#issuecomment-803872517, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALNJVR4AMWV7EKPK2QFYFD3TE36SPANCNFSM4ZORL3BA .

taewookim0812 commented 3 years ago

No I haven't. But if you set the discount factor(DF) to 0 and then train, it would be more close to the deep learning approach.

I don't remember clearly, but when I trained the model with 0 DF, the similarity score is higher than the positive DF(about 0.99) while the overall generality of motion retargeting is worse.

The result may be different in case of pure deep learning and many other variables can affect it.