VicentoWang

Thanks for you work, And I want to ask for advises about styleflow used for image-to-image translation.

1

all I needed is image-to-image translation with attribute-editting function, I first use pix2pix，but I cannot edit it for having no latent-space in pix2pix model. then I plan to try pix2style2pixl...

" The reason is that, if we predict the pose in world space: (different 3d pose ⊕ different camera view) = same 2d projection, then if given a same 2d...

there is a 3d view controle problem.

5

when I change the 3d view of chessboard, there is some kind of lost control of the 3d view rotation, becasue of some unexpected auto rotate from x-axis 

i tend to accout for that #2 means layer2 and x2 means layer2's outputs,is that right ? thankyo

I wonder about what parameter of reverse_augmentation and data_augmentation for ?

1

any difference in training phase or testing phase ？

3D Absolute Human Pose Estimation on Human3.6M Error Defined as MRPE in paper RootNet, could you release the metric MRPE of your Motionet, that is the L2 error on root...

how to achieve realtime object detection + tracking + 2d pose + 3d pose estimetion from a carmera

5

as you demo video demonstrated. https://www.bilibili.com/video/BV18f4y197R7/

MPJEP=43.4 ? really?

how to load this pkl for training ? https://s3.eu-central-1.amazonaws.com/avg-projects/stylegan_xl/models/pokemon512.pkl resume and stem args have some shape problem 

FPS of layer by layer inference is amazing. I wonder how to do that. Thanks