chaitanya100100 / tailornet Goto Github PK

Read blend: /home/arash/Desktop/TailorNet/./visualization/scene.blend
['/usr/lib/blender/blender', '/home/arash/Desktop/TailorNet/visualization/blender_renderer.py', '--', '--body', '/tmp/body.obj', '--gar', '/tmp/gar.obj', '--outpath', '/home/arash/Desktop/TailorNet/outputs/img_0002.png', '--gar_classes', 'pant']
usage: blender [-h] [--outpath OUTPATH] [--body BODY [BODY ...]]
               [--gar GAR [GAR ...]]
               [--gar_classes GAR_CLASSES [GAR_CLASSES ...]]
blender: error: unrecognized arguments: /home/arash/Desktop/TailorNet/visualization/blender_renderer.py -- --body /tmp/body.obj --gar /tmp/gar.obj --outpath /home/arash/Desktop/TailorNet/outputs/img_0002.png --gar_classes pant

Trying to run the render on the predicted garments gives the above error. It seems to throw an error about unrecognised arguments but I'm not sure what exactly it's complaining about. Any help would be appreciated, since this is the final step remaining to achieve inference. Thanks

Hi,

Do you have the dataset for old-t-shirt for male? If have, could you please kindly upload it? Thanks so much.

Thanks for your impressive work, I am very interested in this. However, when I want to have a try and download the dataset, it seems extremely slow, nearly 10KB/s.

So, I come to seek help, I wonder is there any other faster alternative manner such as Google Drive or BaiduYun to acquire the dataset?

What does gamma parm mean? For a garment, is there any way to modify gamma so that the created garment looks like the one we want to create?

How to fit 3D Dress to Human Body from Live Camera ?

any idea about mkaing 3D Dress and fit to live Human body ? share any idea , Model for 3D Dress making and fit to Human Body?
Like Virtual Dress Fit?

Hi, what is needed to put a real garment on an smpl body model? Do I need a 3d model of the garment to do this without losing texture details?

Hi, I run into this error when executing python run_tailornet.py on a remote linux machine:

Traceback (most recent call last):
  File "run_tailornet.py", line 6, in <module>
    from psbody.mesh import Mesh
  File "/data/miniconda3/envs/pytorch/lib/python3.8/site-packages/psbody/mesh/__init__.py", line 10, in <module>
    from .meshviewer import MeshViewer, MeshViewers
  File "/data/miniconda3/envs/pytorch/lib/python3.8/site-packages/psbody/mesh/meshviewer.py", line 49, in <module>
    from OpenGL import GL, GLU, GLUT
  File "/data/miniconda3/envs/pytorch/lib/python3.8/site-packages/OpenGL/GLUT/__init__.py", line 5, in <module>
    from OpenGL.GLUT.fonts import *
  File "/data/miniconda3/envs/pytorch/lib/python3.8/site-packages/OpenGL/GLUT/fonts.py", line 20, in <module>
    p = platform.getGLUTFontPointer( name )
  File "/data/miniconda3/envs/pytorch/lib/python3.8/site-packages/OpenGL/platform/baseplatform.py", line 344, in getGLUTFontPointer
    raise NotImplementedError(
NotImplementedError: Platform does not define a GLUT font retrieval function

Can someone help me with this issue? Thanks!

Hi, I'm just trying to see if I can run this in my enviornment, and I'm getting an error because it's looking for a gamma_000.npy file. Where can I find this file and/or how do I fix this error?

Thanks in advance!

Unable to load font: GLUT_STROKE_ROMAN
Unable to load font: GLUT_STROKE_MONO_ROMAN
Unable to load font: GLUT_BITMAP_9_BY_15
Unable to load font: GLUT_BITMAP_8_BY_13
Unable to load font: GLUT_BITMAP_TIMES_ROMAN_10
Unable to load font: GLUT_BITMAP_TIMES_ROMAN_24
Unable to load font: GLUT_BITMAP_HELVETICA_10
Unable to load font: GLUT_BITMAP_HELVETICA_12
Unable to load font: GLUT_BITMAP_HELVETICA_18
Traceback (most recent call last):
  File "./run_tailornet.py", line 138, in <module>
    run_tailornet()
  File "./run_tailornet.py", line 79, in run_tailornet
    thetas, betas, gammas = get_single_frame_inputs(garment_class, gender)
  File "./run_tailornet.py", line 42, in get_single_frame_inputs
    get_style('000', garment_class=garment_class, gender=gender),
  File "/Users/jihoonkim/Desktop/UCLA/capstone/TailorNet/dataset/canonical_pose_dataset.py", line 11, in get_style
    gammas = np.load(os.path.join(
  File "/Users/jihoonkim/opt/miniconda3/lib/python3.8/site-packages/numpy/lib/npyio.py", line 416, in load
    fid = stack.enter_context(open(os_fspath(file), "rb"))
FileNotFoundError: [Errno 2] No such file or directory: '/Users/jihoonkim/Desktop/UCLA/capstone/TailorNet_dataset/t-shirt_male/style/gamma_000.npy'

Hi,

refer to part 5 dataset generation, is there any detailed instruction on how to generate paired data of parameters and the garment? Thanks.

is there any way to speed up the infrence. It seems it processes a single theta,beta,gamma triplet in 5 seconds with gtx 1070. Is it normal or am I doing something wrong?

I have noticed the closed issue (here), but when I use visualize_two_garments_body function to replace visualize_garment_body in run_tailornet.py with:

visualize_two_garments_body(
      pred_gar, body, os.path.join(OUT_PATH, "img_{:04d}.png".format(i)),
lower_gc='skirt', upper_gc='t-shirt', side='front')

run python run_tailortnet.py render
an error occurred:

Traceback (most recent call last):
  File "run_tailornet.py", line 143, in <module>
    render_images()
  File "run_tailornet.py", line 128, in render_images
    pred_gar, body, os.path.join(OUT_PATH, "img_{:04d}.png".format(i)), lower_gc='skirt', upper_gc='t-shirt', side='front')    
TypeError: visualize_two_garments_body() missing 1 required positional argument: 'outpath'

especially I have tried to use the function get_sequence_inputs, but this error still occurred.
Could you tell me how to fix this problem?
and what`s the pipeline of rendering two_garments?
@chaitanya100100
Thanks!

Hi,

If change the prediction output as .obj files, how would I create the vt and f information inside the .obj file? Please kindly send me examples/templates of each garment if you have them.

Besides, is there a way to change the number of the vertices and faces for prediction output. Thank you so much.

Best Regards,
JC

I ran the code successfully and thanks for the amazing work.
i have 2 questions
section 6.3 in the paper says that we can also transfer the textures from real scan on our predictions. can you suggest a method of doing that?

Also how can i control the garment size? i couldn't figure out how from the paper.

Excuse me if the questions seem naive as i don't have much experience in the field. Thanks in advance

Any suggestion to install blender 2.79? I install 2.91 when I tried then I get an error in the below:
AttributeError: 'RenderSettings' object has no attribute 'alpha_mode'

Process gets killed while loading weights, after running python run_tailornet.py

Gender: male
Cloth style: short-pant

Apparently it gets killed while loading 'tn_orig_ss2g'

The exact same happens in case of t-shirt female.

And it seems like an Out of Memory Error.

Any idea how to fix this?

Hi Author,

Based on the information in the paper, seems there is a model which can be used for generating different style garment mesh based on style-shape. Could you please share this model to us? thanks.

Hi Chaitanya,

I was looking at blender_renderer.py. Can you please add default arguments ?. e.g what should be the input to the following arguments. --body, --gar, --garment_class

Thank you for this wonderful work.

Hi, when I put pose parameters from another model, I get bad results for the right side of the body. How to resolve it?

Hi, I use a human mesh reconstruction tool that infers based on the SMPL neutral model. Is there any way that I can convert the parameters to fit the gendered model(s) that is used by Tailornet?

Hi @chaitanya100100,
Thanks a lot for sharing the code and it makes very interesting step in the subject of dressing up SMPL model. In your codes, we can only render up cloth, do you mind to shed more codes on how to dress the SMPL model completely?

Hi, from the code, it requires the following SMPL models.
SMPL_PATH_NEUTRAL = '/BS/RVH/work/data/smpl_models/neutral/basicModel_neutral_lbs_10_207_0_v1.0.0.pkl'
SMPL_PATH_MALE = '/BS/RVH/work/data/smpl_models/lrotmin/lbs_tj10smooth6_0fixed_normalized/male/model.pkl'
SMPL_PATH_FEMALE = '/BS/RVH/work/data/smpl_models/lrotmin/lbs_tj10smooth6_0fixed_normalized/female/model.pkl'

I download the SMPL_python_v.1.0.0.zip, but inside the folder, only has the male and female models. No neutral model. could you please suggest where to get the neutral one? And from the names above, seems the SMPL Models of male and female have been normalized. Do I understand right, or the models are same as the downloaded ones? thanks.

I download the female and male pkl from smpl website, but when I use pikle load the pkl, will cause error

dd = pkl.load(open(self.get_smpl_file(), 'rb'), encoding='latin1')
_pickle.UnpicklingError: invalid load key, '\x00'.

can you provide the smpl pkl?

Hi,
It may be useful and more convenient for others like me(green hand in this field) if you could please provide the environment requirements in detail such as:

• cuda version
• Pytorch version
• and the version of other libs
  More specifically, maybe a requirments.txt file would be so helpful.

Best,

Hi,

Can i get a sample code for doing multi mesh registration as i am having a hard time figuring how to do that? I am trying to use my 3D scans and hoping to register those to tailornet data.

Thanks for the great work tho.

thanks for your code, greate work!!
i use expose model params to test this code,
when i use the betas, is not right, looks like the shape is different from smplx
the body_pose i change to check the pose is ok
please help

Hi, there are only 4 numbers to modify the garment. How to get a garment from multi garment network as said in oral presentation? I couldn't find any parameter other than gamma to modify the garment.

Hi, it seems you have applied a different strategy for skirts. Did you calculate skinning weights for the skirt with data from PBS based method then train a model or does it include somehow rule-based solution?

DISCLAIMER: I am deploying Tailornet for academic dissertation

1. I have applied a simple dynamic quantization which reduced the model sizes to 1/4. How do I evaluate the accuracy vs original model weights?

2. Applying any sort of PyTorch pruning even with very very small amount results in all weights to be removed. Any idea why this happens?

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

According your your readme, you have mentioned that the program is using random inputs and that we can play with different inputs. But you have not explicitly mentioned what the exact inputs that the TailorNet network takes to make a prediction. For example, can we only use SMPL model to apply garment meshes on the model? Is there a method to use my own model that I created using SMPL model and apply the garment meshes on top of that? Please guide, thanks.

Hi, did you train any model to match garment and body vertices which is the "I" parameter in this equation 3.3 below?
TG(β, θ, D) = I T(β, θ) + D

I want to try my t-shirt bought from a brand using tailornet. Am I supposed to create the rgb image of my t-shirt to some 3d template if so how to do that or I can only try the garments in MGN templates?

Hi,I reproduced your code according to your idea, and the result appeared such an error. What should I do?

And when I follow the steps to train the model, this error will be reported. What should I do?

Is it possible to directly render the result in web based graphic libraries in real time, instead of using Blender?

i am having an issue getting it working with a .ply file output from another model

is it possible to use a .ply file instead of theta , beta ?
if not . can you please guide us ?
thank you.

Hello, thank you for your excellent work. I want to use the "short-pant" clothing model, but found that it is not in the "garment_class_info.pkl" file. Would you mind updating it? Thank you very much!

hi，is it allowed to use a single picture as input?If so, how can I do it?

Hi, I want to try TailorNet on my data.
What is the input format? A 3D obj or a 2D image?
If it is 3D, should I first use something like MGN to produce a 3D obj of my 2D input?
for example, if I want to use the below woman as a body shape input and the coat image as a garment and as a result see the coat on the woman.

@chaitanya100100

Hi Chaitanya,

Can you please share the blender version you used in your work?. I am getting following error
AttributeError: 'RenderSettings' object has no attribute 'alpha_mode'

Does the TailorNet support texture reconstruction? Thanks.

So this happens while using weights & data for male shirt. Female/Male tshirt, Male Short Pants work absolutely fine.

Hi, from the code, it requires the following SMPL models. but cannot find them. could you please upload them? thanks.

SMPL_PATH_NEUTRAL = '/BS/RVH/work/data/smpl_models/neutral/basicModel_neutral_lbs_10_207_0_v1.0.0.pkl'
SMPL_PATH_MALE = '/BS/RVH/work/data/smpl_models/lrotmin/lbs_tj10smooth6_0fixed_normalized/male/model.pkl'
SMPL_PATH_FEMALE = '/BS/RVH/work/data/smpl_models/lrotmin/lbs_tj10smooth6_0fixed_normalized/female/model.pkl'

Thanks for your excellent work.
I download the dataset as you suggested. When I trained the base_trainer, it tells me that some data of t-shirt_male doesn't exist.

../TailorNet_dataset-master/data/t-shirt_male/pose/006_023/unposed_002.npy doesn't exist.
../TailorNet_dataset-master/data/t-shirt_male/pose/006_016/unposed_009.npy doesn't exist.
../TailorNet_dataset-master/data/t-shirt_male/pose/006_016/unposed_013.npy doesn't exist.
../TailorNet_dataset-master/data/t-shirt_male/pose/006_006/unposed_009.npy doesn't exist.
../TailorNet_dataset-master/data/t-shirt_male/pose/006_006/unposed_013.npy doesn't exist.

Well, it finished training anyway. But when I trained the lf_trainer, the same issue came up. And then it just finished without training the lf_trainer. I have no idea of what happened, could you kindly look through my problem. Thanks in advance.

../TailorNet_dataset-master/data/t-shirt_male/pose/006_023/unposed_002.npy doesn't exist.
../TailorNet_dataset-master/data/t-shirt_male/pose/006_016/unposed_009.npy doesn't exist.
../TailorNet_dataset-master/data/t-shirt_male/pose/006_016/unposed_013.npy doesn't exist.
../TailorNet_dataset-master/data/t-shirt_male/pose/006_006/unposed_009.npy doesn't exist.
../TailorNet_dataset-master/data/t-shirt_male/pose/006_006/unposed_013.npy doesn't exist.
Using Smoothing in the dataset
../TailorNet_dataset-master/data/t-shirt_male/pose/006_023/unposed_002.npy doesn't exist.
../TailorNet_dataset-master/data/t-shirt_male/pose/006_016/unposed_009.npy doesn't exist.
../TailorNet_dataset-master/data/t-shirt_male/pose/006_016/unposed_013.npy doesn't exist.
../TailorNet_dataset-master/data/t-shirt_male/pose/006_006/unposed_009.npy doesn't exist.
../TailorNet_dataset-master/data/t-shirt_male/pose/006_006/unposed_013.npy doesn't exist.
../TailorNet_dataset-master/data/t-shirt_male/pose/007_011/smooth_unposed_000.npy doesn't exist.

Great job! The paper reported an experiment of Graph CNN for the style-shape model. It is interesting that MLP outperforms GCN here. I would like to test some ideas based on your GCN network. Would you plan to release its code? It is also good if you can provide detailed network settings. Thanks a lot!

Hi,

could you please advise what is the difference for 'f' and 'ft' for garment mesh and how to get the 'ft'?

for same garment mesh, what should i do to get the .pkl file from .obj file. thanks.

I follow the default setting to train the model, and I found it took long time to load the dataset during training, almost more than five minutes. Is there something wrong?