search

xxlong0 / Wonder3D

Single Image to 3D using Cross-Domain Diffusion for 3D Generation
https://www.xxlong.site/Wonder3D/
GNU Affero General Public License v3.0
4.49k stars 351 forks source link

Fine-tuning Problem #169

Open yufeng9819 opened 2 months ago

yufeng9819 commented 2 months ago

Hey,

Thanks for your wonderful work! I‘d like to ask a question about fine-tuning on other dataset.

Now I follow your rendering script to render another dataset but change the camera pose parameters and rendering angles to get only 4 views (front, back, left and right). I'm wondering if there are any problems with fine-tuning a ckpt based 6-views to get a 4-views model.

Looking forward to your reply!

yufeng9819 commented 2 months ago

Hey,

Sorry to disturb you again. I have another question that I render one dataset following your rendering script to get 9 views images. But I find that the saved information in RT.txt for 9 views is different from your given RT.txt file (./mvdiffusion/data/fixed_poses/nine_views). Is there any problems if I only load fixed poses information rather than load the saved *RT.txt from rendering script.

Looking forward to your reply!

xxlong0 commented 2 months ago

Hello. You may check the rendering code to adjust the number of camera poses, also their values.