search

3DTopia / LGM

[ECCV 2024 Oral] LGM: Large Multi-View Gaussian Model for High-Resolution 3D Content Creation.
https://me.kiui.moe/lgm/
MIT License
1.71k stars 118 forks source link

own data #72

Open shichao1232 opened 3 months ago

shichao1232 commented 3 months ago

How can I convert my images to a format like gobjeverse?

xiyichen commented 3 months ago

I was able to do inference with my own data by converting the camera parameters to opengl convention. I also found it important to crop the images so that the objects are centered, and to resize the images to 256x256.

shichao1232 commented 3 months ago

I was able to do inference with my own data by converting the camera parameters to opengl convention. I also found it important to crop the images so that the objects are centered, and to resize the images to 256x256. I don't konw how to do this ,can you teach me? how can i get the camera parameters? Is it okay to use colmap?

xiyichen commented 3 months ago

Yes, you can use colmap, but it uses opencv convention. I found this function https://kit.kiui.moe/_modules/kiui/cam/#convert helpful for camera coordinates convertion.