VIBE

VIBE copied to clipboard

VIBE is running, but it only uses the CPU (not the GPU) and then crashes as it runs out of memory (64GB..)

5

Hello, I have followed the instruction from Carlos Barreto for installing the VIBE code on a windows PC, and got it to run ... But, it only gets as far...

richardmking

WARNING: You are using a SMPL model, with only 10 shape coefficients. 0%| | 0/2326 [00:00

Oratacth

Thanks for your interest in our research! In amass_utils, it only deals with .npz files. But some datasets in AMASS provides mp4 files, like EyesJapanDataset. How to deal with these...

VioletGreenLi

I really couldn't understand the boxes here

1

The box send to vide is clearly x1y1x2y2.... But why it read as cxcyh? inside this func?? ``` def convert_crop_cam_to_orig_img(cam, bbox, img_width, img_height): ''' Convert predicted camera from cropped image...

lucasjinreal

Hi, I have a customized model, not smpl model, and I want output sketelon alone, how to do that?

luohao123

The scale problems

1

It seems that there are two different scales for processing bounding boxes. In lib/data_utils/img_utils.py, line 135, the scale is 1.2 which seems to be used by Dataset2D/Dataset3D to transform keypoints....

halcyon370

Unstable training when using provided config file --> Early exiting due to MPJPE error

1

Hi all, I'm training VIBE using the provided config.yaml file (i.e datasets, hyperparams, seeds, etc). After the 9th epoch, I get the message: "MPJPE error is 93.05937588214874, higher than 80.0....

grau4

Motion discriminator loss

1

Thanks for your great work! It seems that the l2 disc loss is not bounded (e.g. by a sigmoid function), although in the paper it is stated that the output...

GuyTevet

dockerize VIBE [WIP]

27

x0rzkov

[BUG]demo does not act so well

6

appreciate your work. i find something that i cant explain. when i run the demo on some video taken by my cellphone, it doesnt work well. the reconstruction is always...

sunwenzhang1996