search

Lakonik / MVEdit

3D-Adapter: Geometry-Consistent Multi-View Diffusion for High-Quality 3D Generation
https://lakonik.github.io/3d-adapter
MIT License
281 stars 13 forks source link

Question #19

Open c1ircle opened 1 week ago

c1ircle commented 1 week ago

Where is the inference code of image to mesh in 3DAdapter？

Lakonik commented 6 days ago

Hi! All the inference code is based on Gradio. After starting the webui, you can run these python scripts to start inference.

Note that this repository does not include the GRM Adapter yet (we will have to wait for the official release of GRM unfortunately), so if you run the code locally, you only have access to the MVEdit Adapter for image-to-3D tasks. To do inference with the latest GRM adapter, please refer to this script, which connects to our official demo server for inference.

c1ircle commented 6 days ago

Thank you for your reply. What camera parameters should I use if I want to get images with Gaussian rendering generated by the GRM Adapter? And, is it possible to get 3d gaussian by writing code locally without running gradio？

Lakonik commented 6 days ago

GRM Adapter uses Zero123++ v1.2 as the base model, which is compatible with a wide range of input cameras, so there is no specific requirement. Our Gaussian-related code will be released after the official release of GRM. After that, you will be able to run the demo locally for inference. Non-gradio inference script can be written by calling the methods in https://github.com/Lakonik/MVEdit/blob/main/lib/apis/adapter3d.py, but we do not recommend this since many default parameters are handled by the Gradio interface.