search

AaronJackson / vrn

:man: Code for "Large Pose 3D Face Reconstruction from a Single Image via Direct Volumetric CNN Regression"
http://aaronsplace.co.uk/papers/jackson2017recon/
MIT License
4.52k stars 746 forks source link

About training time.... #93

Closed likewind1993 closed 6 years ago

likewind1993 commented 6 years ago

Thanks for your work! I wanna know about devices and time in detail when you trained the vrn network, For example, you used TITAN GPU and spent one or two weeks train it. These data not shown in the paper, so could you please share it? Thank you again~

AaronJackson commented 6 years ago

Hey,

Two titan 1080 Ti, takes a few days assuming your CPU can keep up with the volumetric augmentation.

Aaron

likewind1993 writes:

Thanks for your work! I wanna know about devices and time in detail when you trained the vrn network, For example, you used TITAN GPU and spent one or two weeks train it. These data not shown in the paper, so could you please share it? Thank you again~

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.

-- Aaron Jackson PhD Student, Computer Vision Laboratory, Uni of Nottingham http://aaronsplace.co.uk

zhoujianwen commented 6 years ago

@likewind1993 Thank you very much for your work. I would like to know how to train volume regression network model. Is there any relevant introductory learning materials?

zhoujianwen commented 6 years ago

@AaronJackson Thank you very much for your work. I would like to know how to train volume regression network model. Is there any relevant introductory learning materials?

likewind1993 commented 6 years ago

@AaronJackson I see, thanks for your reply!

likewind1993 commented 6 years ago

@zhoujianwen I still looking for a way to training it.

AaronJackson commented 6 years ago

Voxelise your dataset and train an hourglass network as described in the paper

likewind1993 commented 6 years ago

@AaronJackson Did you meet the problem of "out of memory", I can't complete calc loss once, it shut down with error of out of memory. The calculated amount is very big(192x192x200=7372800). My computer's memory is 32GB, I don't how to optim the loss function or others, could you give me some advise?