Single 2D image to Metahuman character Demo!
Вставка
- Опубліковано 16 вер 2024
- For Best results:
1. Picture should be high quality.
2. Face expression should be neutral.
3. Click unlit option and change the POV during mesh to metahuman step.
Note: This Open source project uses CUDA enabled graphics card to run. I used RTX 2060 in this tutorial, RTX 30xx series didn't support because of driver issue, I tried a lot with RTX 3070 but couldn't run.
Links:
1. 2D image to 3D Face: github.com/sic...
2. face-alingment package for landmarks detection: github.com/1ad...
3. Python file link for landmark extraction from the input file(place this inside the "Deep3DFaceRecon_pytorch" folder): github.com/Viv...
You are a super professional man!!!!
Oh cool. I never knew about this awesome trick. Thanks for sharing it bro.🔥
Beautiful tutorial!! 💎
2 parts you guys may be missing
pip install packaging
pip install face-alignment
thanks man.. I rewatched the video a few times like an idiot. also not smart not to mention this somewhere in the video..
Please, I need help pip installing face-alignment. Having trouble with it. Thanks /
@kewk thanks for the help, @illusiongraphy my bad man, it was already mentioned in the description and it was clear in the steps of git repository.
@bcrbdb github.com/1adrianb/face-alignment#installation
do this 'after' conda activate deep3d_pytorch
I had been good before I read "RTX 30xx series didn't support" text)
There are lots of bit's missing in this guide. VS C++ SDK is required along with the CUDA toolkit if you want to run this on windows, you will also need to update PyTorch if you have a RTX30x0 or newer.
Amazing bro you're a legend.
bro the installation process is difficult to understand, could you make a separate video to explain it, or a link to clear instructions?
great video btw
Hi, please follow the steps mentioned in github links in the description.
Thank you for this tour.. Nice weekend.
can u give tutorial for setting up deep3d pytorch ? please
😄
Thanks, I have installed face alignment, but why it still shows up 'No module named 'face_alignment'' ?
same question here. I think we have to install the face_alignment package but I don't know where to put it at
I keep getting this "CondaEnvException: Pip failed" even though I've got both conda and pip installed.
Hi Vivek, I’ve been trying to set this up but keep getting errors even though I’ve got python and conda running, also some of the dependencies have been deprecated. Can you please make an updated tutorial?
i was confuse how to install the deep3dface
if i import face fbx to unreal engine it shows me like broken polygons and texture, i dont know about the issue
You're suppose to clean up the mesh before importing into blender, you can clean the mesh in blender .Watch the tutorial by bad studios about how to create a metahuman from a polycam scan in which they also needed to clean the mesh
Hi Vivek very interesting video
Anyone knows this error? File "face.py", line 7, in
preds = fa.get_landmarks_from_directory('input')
File "\anaconda3\envs\deep3d_pytorch\lib\site-packages\face_alignment\api.py", line 288, in get_landmarks_from_directory
dataloader = torch.utils.data.DataLoader(dataset, batch_size=1, shuffle=False, num_workers=2, prefetch_factor=4)
TypeError: __init__() got an unexpected keyword argument 'prefetch_factor'
Epic
i have problem with default photo not changing to whatever picture i choose, also i have radeon rx gpu, can it work?
Sorry I don't think it will work. It's dependant on Nvidia's CUDA.
wow that great
Can you put this project on google collab and make it available?
bro can i make medusa queen chracter in this system chracter allready 3d
It is possible in theory. Please try and uppload your video!
good job bro.... 😎
Great video! Thanks. Just emailed you regarding opportunity.
Hey, I am trying my luck with an RTX 3080, can you tell me where exactly it failed? I am getting as far as "loading the model from ./checkpoints\model\epoch_20.pth" but it seems to hang there
for me it definitely throwed error related to device not supporting. Basically, lower version of pytorch needed to run this code properly that version doesn't support latest cards. sorry for the delayed response.
Process and Idea is great.. results however are not superior than manually editing Metahuman IMHO..
Okay!
the whole point is to give you a head start. You have to adjust it to match your needs
You need to do another tutorial for this, this one lack explanation and it's impossible to understand
Sorry for the late reply. Hope you have upgraded to latest versions that enables Metahumans creation directly from captured video!
Hello, I have a question about how to install the "face-alingment package for landmarks detection". Is it necessary to invoke the "face.py" script?
Yes! please follow the instructions from the link in description
cd Deep3DFaceRecon_pytorch
pip install face-alignment
you know by any chance how to import metahuman with the animation and everything textures, hair ect. In to a 3D Program like blender or similar to then render it there?
Export metahuman from UE5 in fbx format I guess.
for the blender, after that, you must definitely use the fbx converter so as not to lose morphs
Awesome yarr 😲😲😲
hello is this an error. Please guide me to fix it. I use windows 11 and rtx3090 "
E:\PM 3D FACE\Deep3DFaceRecon_pytorch>conda activate deep3d_pytorch
'conda' is not recognized as an internal or external command,
operable program or batch file."
Hi, it looks like you have not installed anaconda. You need python developing knowledge for this method! Sorry!
@@vivekanandhan555 Bro, not all of us are developer. If we need python developing knowledge, so why you are making a step by step tutorial video? If anyone cannot make this without python developing knowledge, this video is useless then.
is there any easier steps than thiss
But still likeness is not accurate ..
You need artist to do more work on likness
Agreed!
Hi! I'm getting this error when trying to run face.py
Traceback (most recent call last):
File "face.py", line 7, in
preds = fa.get_landmarks_from_directory('input')
File "C:\Users\erik8\anaconda3\envs\deep3d_pytorch\lib\site-packages\face_alignment\api.py", line 288, in get_landmarks_from_directory
dataloader = torch.utils.data.DataLoader(dataset, batch_size=1, shuffle=False, num_workers=2, prefetch_factor=4)
TypeError: __init__() got an unexpected keyword argument 'prefetch_factor'
How do I fix it?
Hi, I'm guessing you have installed a different pytorch version. Not sure though.
@@vivekanandhan555 what version are you using in the video?
If you create python environment using this file github.com/sicxu/Deep3DFaceRecon_pytorch/blob/master/environment.yml, version should be taken care automatically. In my machine, these are the version:
tensorflow==1.15.0
torch==1.6.0
@@vivekanandhan555 I’ll try it! Thank you so much
@@erikluo5006 I got the same error. Were you able to fix it?
indian dudes finna be making thier own corn now with nood skins lmao
🤣
Does it work with RTX4070Ti?
I guess so, please try and let as know.
But we'll tried
sorry?
one tutorial in spanish please
Not very useful without more blendshapes.
Okay
ua-cam.com/video/sxJoWZ0qBcE/v-deo.html
i make a tutorial to achieve that an image was converted to metahuman using a new method
я тебе че айти специалист???
I'm a software engineer yes!
It looks nothing like her.
okay
Bro I need one character give me your email address