search

yw0208 / W-HMR

W-HMR: Monocular Human Mesh Recovery in World Space with Weak-Supervised Calibration
https://yw0208.github.io/w-hmr/
MIT License
30 stars 0 forks source link

Unable to load EGL library #7

Open harikrik007 opened 4 months ago

harikrik007 commented 4 months ago

I am using window11 , anaconda environment with python 3.8.

utils\renderer_cam.py", line 130, in render_overlay_image renderer = pyrender.OffscreenRenderer(

The above line produces an exception

OpenGL\platform\egl.py", line 86, in EGL raise ImportError("Unable to load EGL library", *err.args) ImportError: ('Unable to load EGL library', "Could not find module 'EGL' (or one of its dependencies). Try using the full path with constructor syntax.", 'EGL', None) Installed pyopengl,pyopengl-accelerate

I am able to get pkl file with --no_render param but i need the .obj file Please help me rectify the issue

yw0208 commented 4 months ago

OK, you have saved SMPL params in pkl files, then you can get obj files like following codes: ` smpl->vertices->mesh->obj

output = smpl_model(body_pose=rotations, global_orient=global_orient, betas=betas)
vertices = output["vertices"]
# visualize: vertices to obj
mesh=Trimesh(vertices=vertices,faces=smpl_model.faces)
save_path="scratch/temp/mesh_obj/mesh.obj"
with open(save_path, 'w') as fw:
    mesh.export(fw, 'obj')`
harikrik007 commented 2 months ago

Thanks for your help. 1) I have tried out the inference and got the output meshes. It looks cool. The issue that I found was that, visual accuracy and measurement accuracy is lagging for a case of a plus size person. Would like to hear your comments on this to know if I missed anything. 2) Meanwhile I saw your mention about the training codes, would like to see it when it’s ready. Best Regards,

yw0208 commented 2 months ago

Thanks for your help.

  1. I have tried out the inference and got the output meshes. It looks cool. The issue that I found was that, visual accuracy and measurement accuracy is lagging for a case of a plus size person. Would like to hear your comments on this to know if I missed anything.
  2. Meanwhile I saw your mention about the training codes, would like to see it when it’s ready. Best Regards,

Hi, I have noticed your questions, here are my reply.

  1. A plus size person? Do you mean these overweight humans? I think this is mainly due to the lack of training data. Although I have used most of the classic datasets, they are relatively lacking in samples of overweight people. For this kind of overweight people, you can refer to this article Everybody Is Unique: Towards Unbiased Human Mesh Recovery. I remember there are other papers that try to solve this problem, but I forget the paper names. You can try google them.
  2. I had planned to make all the code public after the paper was published, but it seems that this will take a long time. So I plan to make all the code public next week. However, the code has not been organized, so it may be confusing and have bugs. Please feel free to contact me when the time comes.