AI Generates 3D Model From Single Photo (Google Colab)
Vložit
- čas přidán 17. 05. 2021
- This AI can generate a 3D model from a single photo. One picture and you have a complete 3d model of the person. You could rig this model, animate them, 3d print it and do all kinds of cool stuff. This project uses Google Colab free gpu, so anyone can render this project without a powerful computer.
Colab Notebook: tinyurl.com/4b...
Man Picture: i.pinimg.com/o...
Original Creator: • PIFuHD: AI constructs ...
I hope this video saved you alot of time from modeling a figure yourself :) ganverse3d
Instructions unclear, I scanned myself into the computer and now I cant get out
😂
💀
Beo this comment Got me crying💀💀💀
Do you have a Digimon partner? 😂
😂
No longer works. ERROR: Could not find a version that satisfies the requirement pytorch3d (from versions: none)
its not working
ERROR: Could not find a version that satisfies the requirement torch==1.6.0+cu101 (from versions: 1.11.0, 1.11.0+cpu, 1.11.0+cu102, 1.11.0+cu113, 1.11.0+cu115, 1.11.0+rocm4.3.1, 1.11.0+rocm4.5.2, 1.12.0, 1.12.0+cpu, 1.12.0+cu102, 1.12.0+cu113, 1.12.0+cu116, 1.12.0+rocm5.0, 1.12.0+rocm5.1.1, 1.12.1, 1.12.1+cpu, 1.12.1+cu102, 1.12.1+cu113, 1.12.1+cu116, 1.12.1+rocm5.0, 1.12.1+rocm5.1.1, 1.13.0, 1.13.0+cpu, 1.13.0+cu116, 1.13.0+cu117, 1.13.0+cu117.with.pypi.cudnn, 1.13.0+rocm5.1.1, 1.13.0+rocm5.2, 1.13.1, 1.13.1+cpu, 1.13.1+cu116, 1.13.1+cu117, 1.13.1+cu117.with.pypi.cudnn, 1.13.1+rocm5.1.1, 1.13.1+rocm5.2, 2.0.0, 2.0.0+cpu, 2.0.0+cpu.cxx11.abi, 2.0.0+cu117, 2.0.0+cu117.with.pypi.cudnn, 2.0.0+cu118, 2.0.0+rocm5.3, 2.0.0+rocm5.4.2, 2.0.1, 2.0.1+cpu, 2.0.1+cpu.cxx11.abi, 2.0.1+cu117, 2.0.1+cu117.with.pypi.cudnn, 2.0.1+cu118, 2.0.1+rocm5.3, 2.0.1+rocm5.4.2)
ERROR: No matching distribution found for torch==1.6.0+cu101
ERROR: Could not find a version that satisfies the requirement torchvision==0.7.0+cu101 (from versions: 0.1.6, 0.1.7, 0.1.8, 0.1.9, 0.2.0, 0.2.1, 0.2.2, 0.2.2.post2, 0.2.2.post3, 0.12.0, 0.12.0+cpu, 0.12.0+cu102, 0.12.0+cu113, 0.12.0+cu115, 0.12.0+rocm4.3.1, 0.12.0+rocm4.5.2, 0.13.0, 0.13.0+cpu, 0.13.0+cu102, 0.13.0+cu113, 0.13.0+cu116, 0.13.0+rocm5.0, 0.13.0+rocm5.1.1, 0.13.1, 0.13.1+cpu, 0.13.1+cu102, 0.13.1+cu113, 0.13.1+cu116, 0.13.1+rocm5.0, 0.13.1+rocm5.1.1, 0.14.0, 0.14.0+cpu, 0.14.0+cu116, 0.14.0+cu117, 0.14.0+rocm5.1.1, 0.14.0+rocm5.2, 0.14.1, 0.14.1+cpu, 0.14.1+cu116, 0.14.1+cu117, 0.14.1+rocm5.1.1, 0.14.1+rocm5.2, 0.15.0, 0.15.0+cpu, 0.15.0+cu117, 0.15.0+cu118, 0.15.0+rocm5.3, 0.15.0+rocm5.4.2, 0.15.1, 0.15.1+cpu, 0.15.1+cu117, 0.15.1+cu118, 0.15.1+rocm5.3, 0.15.1+rocm5.4.2, 0.15.2, 0.15.2+cpu, 0.15.2+cu117, 0.15.2+cu118, 0.15.2+rocm5.3, 0.15.2+rocm5.4.2)
ERROR: No matching distribution found for torchvision==0.7.0+cu101
ERROR: Could not find a version that satisfies the requirement pytorch3d==0.2.5 (from versions: none)
ERROR: No matching distribution found for pytorch3d==0.2.5
> Searches "AI Modeling" in curiosity
> Finds this video
> Flashback to AI Art
Welp, shit's gonna hit the fan again!
I can say this video is not only worthy but a Gem!!!
no info about texturing
error: OpenCV(4.6.0) /io/opencv/modules/imgcodecs/src/loadsave.cpp:801: error: (-215:Assertion failed) !_img.empty() in function 'imwrite'
Yeah I like that part too
What a pity! An error occurs when configuring step1:
"ERROR: Could not find a version that satisfies the requirement torch==1.6.0+cu101 (from versions: 1.11.0, 1.11.0+cpu, 1.11.0+cu102, 1.11.0+cu113, 1.11.0+cu115, 1.11.0+rocm4.3.1, 1.11.0+rocm4.5.2, 1.12.0, 1.12.0+cpu, 1.12.0+cu102, 1.12.0+cu113, 1.12.0+cu116, 1.12.0+rocm5.0, 1.12.0+rocm5.1.1, 1.12.1, 1.12.1+cpu, 1.12.1+cu102, 1.12.1+cu113, 1.12.1+cu116, 1.12.1+rocm5.0, 1.12.1+rocm5.1.1, 1.13.0, 1.13.0+cpu, 1.13.0+cu116, 1.13.0+cu117, 1.13.0+cu117.with.pypi.cudnn, 1.13.0+rocm5.1.1, 1.13.0+rocm5.2, 1.13.1, 1.13.1+cpu, 1.13.1+cu116, 1.13.1+cu117, 1.13.1+cu117.with.pypi.cudnn, 1.13.1+rocm5.1.1, 1.13.1+rocm5.2, 2.0.0, 2.0.0+cpu, 2.0.0+cpu.cxx11.abi, 2.0.0+cu117, 2.0.0+cu117.with.pypi.cudnn, 2.0.0+cu118, 2.0.0+rocm5.3, 2.0.0+rocm5.4.2, 2.0.1, 2.0.1+cpu, 2.0.1+cpu.cxx11.abi, 2.0.1+cu117, 2.0.1+cu117.with.pypi.cudnn, 2.0.1+cu118, 2.0.1+rocm5.3, 2.0.1+rocm5.4.2)
ERROR: No matching distribution found for torch==1.6.0+cu101"
mine too
me too
not working, pip dependency error as everyone has posted
greta for blocking out a character and making sure the proportions are right, but then you can add to it in zbrush etc
ERROR: Could not find a version that satisfies the requirement pytorch3d==0.2.5 (from versions: none)
ERROR: No matching distribution found for pytorch3d==0.2.5
same here
same
Same
its dead
sucks . . .
Awesome. If it actually worked.
Someone should make something like this some day maybe.
I cant run it, do I need to buy any liciense?
Yeah
error when importing image, i did the copy image address and it still doesnt work, the image im using doesnt exist in pinterest
Same here
Cannot use web address, must be image address
@@Goose_Bomb Don't work anyway. Nor for imageshack nor for imgur.
Keep getting a error line even when I follow the steps it doesn't do it
Is this still working or its my computer problem ???
No longer working, the first step gives an error "ERROR: No matching distribution found for pytorch3d==0.2.5" among other errors
Does this work with an illustration of a character? Or does it only work with photos?
Maybe it's too late, but I watched recently that you can actually do it with an illustration, I don't remember if it was with this same website, but I'm sure you can find a tutorial or demostration very easy. Good luck
As of February 2023 the script doesn't work. It cannot create necessary directories to even upload the images.
I am glad I read the comments .
try scrolling down and running the code block below which allows you to upload image from your system. That worked for me. Also I just checked and realized that this technology is actually from facebook/meta research. It is working reasonably well for images with a clear white background
The first cell doesn't run. Says something about pytorch not found. Pip install pytorch doesn't work either. Cool idea though. Wish it worked.
i just wanna know how the ai knew what the back of the guy looked like
AI learning is fucking crazy. perhaps something like it recognized the front facing human male, went online and used it to referenced millions of side and back images with similar profile and extrapolated.
@@jdogsful fr
@@prod-renxy the craziest thing is, now there are millions of AI generated artworks online created in the past few months that it is also referencing, so it is using AI art to get better at generating AI art, and that is why it is becoming so good so fast. And its only just begun.
Imo 3d artists would feed the AI with full 3d models, enough so that the AI learns from them and can fill the missing information from the database
@@katherinet4769 ohhh
You need to use the old GPU, new GPU not support (old mesh structure). Minimun 8GB-VRAM, Nvidia GPU display card very long (Mini not easy to get), High power supply unit not easy to get. Google-Colab using new GPU. Only work for old version of LINUX, like Ubuntu or POP-OS. New Ubuntu GPU driver not support old GPU. Python 3.7 working, Python 3.10 not working because it support new GPU. Don't try Windows 7/8/10 it is not working and complicated.
error, error, error, error, error.....and so on. Not working like in video. Tried with image links and upload
ERROR: Could not find a version that satisfies the requirement torch==1.6.0+cu101 (from versions: 1.11.0, 1.11.0+cpu,
i keep getting errors while i did the exact same thing i did as you
i think it only works for human 3d models
Keeps giving me an error.. any suggestions? error: OpenCV(4.6.0) /io/opencv/modules/imgcodecs/src/loadsave.cpp:801: error: (-215:Assertion failed) !_img.empty() in function 'imwrite'
same, did u manage to fix it?
@@odesamusic not yet.. Did you find a solution?
@@juancastro6056 did u find a solution?
@@juancastro6056 try scrolling down and running the code block below which allows you to upload image from your system. That worked for me. Also I just checked and realized that this technology is actually from facebook/meta research. It is working reasonably well for images with a clear white background
I really miss hearing this TTS voice on almost every new video. Now they all sound like Josh from Let's Game it Out.
Sadly this does not work, either with an image link or with an uploaded photo. All errors return to "upgrade to pro".
not working: AttributeError: 'NoneType' object has no attribute 'shape'
mkdir: cannot create directory ‘/content/pifuhd/sample_images’: No such file or directory
/content/pifuhd/sample_images/webImg: No such file or directory
----------------------------------------------------------------
Thanks for hosting your model, but the model no longer works. Please help
Great Work !!
WHAT! HOW! how does ! ? wow this is just incredible!
tried it 5 times now and it just wont work. THANKS FOR NOTHING!!!
I'm always thankful. There are clothing photos from various angles. Is there a way to naturally change the clothes of the character in the video?
Amazing👌
Thanks, but... what about the texture?
Great job👏👏
TRY AND COMMENT ..! IF IT WORKS PLS LET ME SOLVE THE ERROR AT 2nd STEP WHILE UPLOADING IMAGE
doesn't work, error always
i knock my head 3x this awesome
I keep getting errors everytime I try to import an image URL
Copy the image address, not the website link 😄
you need to run step one first before uploading you're model. after step one is installed then you're picture will upload without errors
i find it best to get the pictures from Pinterest.
@@CGSheep same error with image address
This does not work
How come? I wanna try it
doesn't work for me. I get the error that it cannot resolve the image url, even after trying several different images on different web sites.
UPDATE: I use the default PiFuHD demo colab notebook and use the one section to upload my own image, and it does work well.
here's a timestamp of a vid where I show me doing this:
czcams.com/video/vIrVuF55rHE/video.html
Wait how?
Walk us through it
how?
@@user-oo6us5nf1ni show how I do it here: czcams.com/video/vIrVuF55rHE/video.html:
here is a timestamp of a video where I show myself doing this:
czcams.com/video/vIrVuF55rHE/video.html
Man AI is so useful
God thanks man, this saves alot of time.
oh i can not, pls help me
error: OpenCV(4.6.0) /io/opencv/modules/imgcodecs/src/loadsave.cpp:801: error: (-215:Assertion failed) !_img.empty() in function 'imwrite'
Can you explain step by step how to use Icon?
It actually worked for you????
@@VArt1998 Idk, I just load an image in and then it happens.
Using a potato laptop also btw.
Not workyng anymore, shows me an error
I almost feel bad for getting this info for free
When I execute STEP 1 in the final part it tells me: ERROR: Could not find a version that satisfies the requirement.
because?
Step 2. Resolving $image_url ($image_url)... failed: Name or service not known. I've tried 3 different locations
yes i also tried no use
Directly run the next cell skipping that one and it will ask u to upload your image and you are done
@@starksharma8612 ok thanks
@@svdevi9912 you are welcome
@Jesus you are welcome bro
love you !!!!
I created monstrosities maybe i have but images. But it worked.
Error : pip dependency conflict
How do you generate the texture for that 3D model.
have to draw it in Blender
@@odesamusic You could do that and tbh its more efficent but you can use stuff like midjourney to create textures for you
@@odesamusic texture painting is an even better alternative
If u cant find an automated way i recommend substance painter as it seems very intuitive and easier than others.
If it will be seen from one angle only, than project the original image onto it
Dang. Too bad it can only do that one guy on Pinterest.
Superb!
Doesn't work for me, I get an error all the time
same
this is not a scam but it is not ai. you pay the site who has human modelers generate your image. its good but freakishly expensive. basically $20 for the lowest res model. Would be worth it for $0.99.
Image address won't work. Copying straight from google. Did the code expire..? Please lmk if i'm not doing it right-
My image is in Google Drive and it fails
It's not 100% good, and when you try to make changes to the model, it doesn't work. It doesn't have the correct mesh.
awesome!
Resolving $image_url ($image_url)... failed: Name or service not known.
wget: unable to resolve host address ‘$image_url’
The project got abandoned guys,its not working
wao, thank you
I'll try it when I can do it offline on my own gpu. I don't do google.
Running the first step leads to errors and it not working.
2024 : ITS NOT WORKING...
What about the texture?
How can you ad your photos , not from a website, but your own photos , any ideas ?
my 1st step keeps coming up with an error what do i need to do
and how about add color after import
and how to move with rigging
Can you use multiple images to better train the A.I?
No, you can just fix the art yourself... Or you can't do that? Get a actual skill and learn sculpting in blender to better enhance the image
@@raesdzaky2757 The whole point of AI generated models is to bypass the sculpting/unwrapping/texturing process all together, which is what small studios need, especially in game development. You can bypass the artist either for prototyping or to produce non AAA games but you can't bypass game designers. Too bad.
@@GhtPTR still annoying AF, either way, its still gonna function bad
It doesn't work anymore, maybe it never did
guess its out of date could not get certain pkgs to install
Can we do this same for the image sequence? for a complete video?
We can use any picture or just for a character
when I launch the first step I get a bunch of errors because I don't have torch, torchvision or pytorch 3d. but how I install them ?
Got it?
Nop
Fail - ERROR: Could not find a version that satisfies the requirement torch...
I think we can't use it now unless we update the code or waiting for someone to update it
I didn't work, I got a random shape
does it works with figures that does not depict humans?
Only one way to find out dude.
No. This AI model was trained to work for people who work in textile.
@@amartious2123 thank you, really useful
Is there a high Poly count on that model?
this no longer works. are there valid alternatives?>
not loading image
can i put it into 3dsmax
Yep this doesn't look like I'll get a virus at all
its not working
the colab dont let me hit connect
😳 Daaaamn
OH NICE. 😮
I hate how I finally finished making a base mesh of my brother for a future game project of mine in nearly 3 months, and this thing took a random guy and did the mesh in a minute.
Now I’m questioning my dream of becoming a video game designer. I even had 4 different ideas for stories and lore.
My bro was literally the perfect model for being the main character in one of them.
I’m just so pissed and disappointed at the same time.
no, no, think the opposite. if you have ideas for stories and lore, now you are gonna have the tools to create them faster, closer to the speed of imagination. you won't have to get bogged down in learning technical skills that AI will soon make obsolete. now you can focus more on the storytelling. that's what I'm doing. these ai tools are allowing me to create a game as rich as skyrim with #NotSSgame. I use Stable Diffusion with ControlNet to create turnaround character textures in a single image (front, side back). even if I model the model, the texture is already well-started, and now I'm hunting for a tool like this to convert the front image into a 3d model, which I can retopo using quad remesher with some loop cut cleanup, then use Accurig to create blendshapes automatically. since I have hundreds of characters to make, having a smooth and fast workflow for this technical stuff, is good.
don't be bummed the tools are better now. just start using them!
@@westingtyler1 Exactly
Also game ready assets are a whole different being then just making a mesh. You gotta optimize it etc. and Ai still sucks at that.
@@jubb1984 yeah totally. i'm actually doing this right now. I'm using PIFuHD to generater 90k triangle untextured models to get into Blender. but the topology is not good. I'm looking into other time-saving tools like Quad Remesher and Softwrap for Blender to make retopo simpler, and FaceIt to generate the blend shapes, but there is still a lot of skill and judgment required to get good animating models this may change in a couple years once the AI is better, but the storytelling part is still gonna be up to us.
@@westingtyler1 yeah its what i ment, also im 100% on board the train that ai will not replace people but finally the mind numbing parts of our work
That's a great idea . Never give up on your dream .
can't get pytorch3d up to date on it
make a video same topic but 2024
Its not working,the steps are not the same.
Why not just use character creator that way its rigged for animation.
good
nice
Modify the following lines of step 2
change this
!wget '$image_url' -O '$path'
For this
!wget $image_url -O $path
and then this
!rm -f '$path'
For this
!rm -f $path
i subscribed after this ....
whaat about the textures?
Texturing isn't as hard as modeling so you should be fine also if you want quick texture but its not that detailed you could use the same photo and in uv editing "project from view" facing the 3d model it'll fit right in and adjust it if you need
@@stylie473joker5 texturing is much harder
@@shadowbaconwarrior5604 depends what type of texturing though cause the more details the harder it gets and the more time consuming
@@stylie473joker5 well actually considering AI Generated Textures is quite common now I could agree that modeling is harder right now. That is until AI can do that too
Ai can also do high quality Animation, Coding,and UI
Even sound design isn't completely safe it's getting harder and harder to be a developer the more the world advances
@@shadowbaconwarrior5604 Yeah its crazy, i still think its years away from perfection but its scary to think what Ai can achieve
upload your photo. no spying...
Can you do it with texture
It is works for his texture too?