zhizdev / mvdfusion

[CVPR 2024] MVD-Fusion: Single-view 3D via Depth-consistent Multi-view Generation
https://mvd-fusion.github.io/
MIT License
106 stars 4 forks source link

the value of `feed_prev_depth` in `mvd_train.config`? #6

Closed thucz closed 5 months ago

thucz commented 5 months ago

I want to train MVD-Fusion on my customed dataset. Can you also provide the mvd_train.config file(including the value of feed_prev_depth)?

zhizdev commented 5 months ago

thanks for checking out or repo!

feed_pred_depth is false during training and optional (default to false) during inference. training code and config are uploaded.

thucz commented 5 months ago

Thanks!