threestudio-project / threestudio

A unified framework for 3D content generation.
Apache License 2.0
6.32k stars 480 forks source link

Magic123 #282

Closed bennyguo closed 1 year ago

bennyguo commented 1 year ago

Magic123 coarse stage implementation with the new unified zero123 and SD guidance.

bennyguo commented 1 year ago

catstatue_rgba

https://github.com/threestudio-project/threestudio/assets/19284678/bf9f6772-93ae-4385-b6b1-46ccd662068f

dog1

https://github.com/threestudio-project/threestudio/assets/19284678/a721eb06-a64b-4040-b7d3-14522e6ee26d

grootplant_rgba

https://github.com/threestudio-project/threestudio/assets/19284678/9ac0d4b1-9015-41d0-ba0f-f8c9674177b4

yankeesong commented 1 year ago

Great work! I tried this simple prompt:

python launch.py --config configs/magic123.yaml --train --gpu 0 data.image_path=/path/to/dragon_rgba.png system.renderer.num_samples_per_ray=256 data.width=64 data.height=64

but got the following error:

MissingMandatoryValue: Missing mandatory value: tag
    full_key: tag
    object_type=dict

Do you have an idea how to resolve this? Thanks!

bennyguo commented 1 year ago

@yankeesong you also need to specify system.prompt_processor.prompt

yankeesong commented 1 year ago

Oh got it. So currently textual inversion is not supported right?

bennyguo commented 1 year ago

Yes and I don't plan to support it :)

ajhamdi commented 1 year ago

Thanks for integrating our Magic123 into threestudio !