zhuhao-nju / hmd Goto Github PK

Hello,

thank you so much for uploading your code.
The link to the pretrained model here seems to be broken. Any idea why?

Is there any available training code of vertex deformation network? It seems it is not included in this repo. Also, pre-trained vertex deformation network is not included as well.

Hi,
I want to test my own data,but I'm wondering how can I get the .json file in the "/test/para" folder?I seem to not find the approach to get the .json file through HMR.

Thanks

Hi, thanks for sharing code of your paper.
Could you share the Kinect depth map dataset used to train Shading-Net?

Hello, thanks for ur work!

I plan to run the datasets/proc_all_data.py to get all the dataset.
But I encounter the issue:

ubuntu@server~/Desktop/hmd/datasets$ python proc_all_data.py
Traceback (most recent call last):
  File "proc_all_data.py", line 6, in <module>
    from LSP import proc_lsp
  File "/home/ubuntu/Desktop/hmd/datasets/LSP.py", line 14, in <module>
    from hmr_predictor import hmr_predictor
  File "/home/ubuntu/Desktop/hmd/datasets/../src/hmr_predictor.py", line 26, in <module>
    import src.config
ModuleNotFoundError: No module named 'src.config'

Is there any missing file(config) in src folder?
(I have already used pip install src)

Thanks in advance : )

Hello,
Ur work seems to be great Is there any upadte on the ttraining code ? @zhuhao-nju @zhuhao-nju

Is it possible to get obj files for the results predicted and looking for how to use demo outside test images

It seems HMD can only produce detailed shape at visible area. Can it produce detailed geometry at invisible area from the image?

Hi! Excellent Work!

I'm interested in your dataset and I wonder how I can get the smoothD_xx data in the dataset. I know that you use some shading refinement techniques from your paper. But could you provide some details about this?
Thanks!

Hi,
I am trying to run the code on the CLOTH3D dataset. I ran the data preprocessing scripts (modified proc_all_data.py script). When trying to create a custom joint dataloader based on the dataloader_joint class, I was stuck on what this meant on line 70

# make target para
tgt_para = np.array(joint_move[(joint_id*2):(joint_id*2+2)])

For every input image, I have saved the joint_move arr but how would I filter it using joint_id for my dataset?

Hello,
I am trying to use an smpl model from smplify instead of hmr for the input and I am trying to understand how to do it. I am using the demo.py and it seems that i need only vertices and vertex normal as input, am I wrong? Because I can't get a solution, especially in the part where you get the silhouette.

Hi,

Thanks for sharing the code.

May I know how could we train the model i.e. could you provide training file?

Thanks

I tried to setup the data in wild demo and get this error after setting up hmr
!python ./predict_hmr.py --img ./input/Duncan.jpg --crop_x 100 100
Namespace(crop_x=[100, 100], crop_y=-1, img='./input/Duncan.jpg', outf='./result/Duncan/', pad=0)
Traceback (most recent call last):
File "./predict_hmr.py", line 48, in
margin = 30, normalize = True)
TypeError: preproc_img() got an unexpected keyword argument 'margin'

Hello, thank you for sharing the great work!

I've found that URL for the pretrained model (http://cite.nju.edu.cn/Haozhu/hmd_pretrained_model.tar.gz) in https://github.com/zhuhao-nju/hmd/blob/master/demo/download_pretrained_model.sh is no longer accessible and it is redirected to the top page of your lab.

FYI: also, [Project Page] link in the README.md is outdated.

It would be great if you could give us a new URL for the pretrained model.
Thanks in advance.

Hello, thank you so much for uploading your code. The link to the pretrained model here seems to be broken. Any idea why?

Thx for your great works.
I wonder if this repo can get joint3d,joint2d and theta like hmr?Or you have plan to do it?

Thank you for sharing your great work.

Could you please share the depth dataset created with the kinetic sensor?

There is a huge improvement in Detailed Avatar Recovery from Single Image, could you share the latest code for research purpose?