search

EricGuo5513 / text-to-motion

Official implementation for "Generating Diverse and Natural 3D Human Motions from Texts (CVPR2022)."
MIT License
462 stars 40 forks source link

About the confidence of generated results #24

Closed lzyplayer closed 1 year ago

lzyplayer commented 1 year ago

Thanks for such great work! Is there a numerical value that can represent the confidence of the generated result pose motion? And It would be great if you could explain the return value. For example, the function generate in trainers.py 384 returns fake_motions, mus_pri, att_wgts. Could you please explain them a little bit?

EricGuo5513 commented 1 year ago

Hi, not sure about the confidence you refer to. Do you mean the reported confidence value in our paper? We calculate the 95% confidence in https://github.com/EricGuo5513/text-to-motion/blob/main/final_evaluations.py#:~:text=def-,get_metric_statistics,-(values)%3A.

In generate function, fake motions are exactly the generated motions, mus_pri and attwgts are only for analysis during the inference process.

lzyplayer commented 1 year ago

Thanks for your help! Then I assume that the motion sequences generated multiple times are equally loyal to the input in the algorithm, am I right?

EricGuo5513 commented 1 year ago

Yes, it is correct.

On Thu, Apr 6, 2023 at 2:41 AM vicky Lin @.***> wrote:

Thanks for your help! Then I assume that the motion sequences generated multiple times are equally loyal to the input in the algorithm, am I right?

— Reply to this email directly, view it on GitHub https://github.com/EricGuo5513/text-to-motion/issues/24#issuecomment-1498697099, or unsubscribe https://github.com/notifications/unsubscribe-auth/AKRYNB26OG24PIQQ2PQ7RZLW7Z6SVANCNFSM6AAAAAAWLUSWZE . You are receiving this because you commented.Message ID: @.***>