search

chaitanya100100 / TailorNet

Code for our CVPR 2020 (ORAL) paper - TailorNet: Predicting Clothing in 3D as a Function of Human Pose, Shape and Garment Style.
https://virtualhumans.mpi-inf.mpg.de/tailornet/
Other
408 stars 67 forks source link

tailornet dataset #28

Closed defunct-ff closed 3 years ago

defunct-ff commented 3 years ago

Why do we need to download the large datasets? The model weights are already provided in this repo's readme.

chaitanya100100 commented 3 years ago

Hello @mfikryrizal .

TailorNet model requires betas (body shape parameters) and gammas (garment style parameters) from the dataset. That is why dataset download is required.

In our dataset organization, we haven't yet separated those parameters from the 3D garment groundtruth(which is responsible for most of storage space) i.e. everything is provided together. I have transferred your query to @zycliao who handles dataset repository. Till then, dataset download will be required for running inference.

Thanks.

defunct-ff commented 3 years ago

Thank you for the response @chaitanya100100 . I had another look at the code (going through run_tailornet.py) and I noticed that pose dataset files -the largest files- are not used in instantiating any model. So is it safe to say that I can just remove them if I use my own SMPL parameters, while shape and style_shape must be retained?

zycliao commented 3 years ago

Thank you for the response @chaitanya100100 . I had another look at the code (going through run_tailornet.py) and I noticed that pose dataset files -the largest files- are not used in instantiating any model. So is it safe to say that I can just remove them if I use my own SMPL parameters, while shape and style_shape must be retained?

Yes. pose is only for training, so you can remove them for just inference. I will pack files that are used for inference later.

defunct-ff commented 3 years ago

@zycliao That's great! Thank you for the response. I'm done with the questions