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

Question about train_pred_image and eval_pred_image at finetune time #44

Closed yisuanwang closed 8 months ago

yisuanwang commented 8 months ago

Thank you for your wonderful work! However, I have a question about the fact that in finetune, both train_pred_image and eval_pred_image only have output on the far left side; the right side is blank. But both GT images are normal 8-view plots. Is this normal? If it's abnormal which place could be the problem? 88d94ceb9146320c4dc5cbf84930c04 4bb7ef1aa4d7d60af7f98de7422d424

ashawkey commented 8 months ago

As answered in the email, this may be caused by wrong camera poses.