zero123plus
zero123plus copied to clipboard
SUDO-AI-3D
Will the training code be published?
I would like to make a model by having it learn in an original camera pose. Hopefully the training code will be made available.
The training code is very specialized towards our infrastructure, and thus we do not have a plan to release a complete training script for now. But every module is there in the diffusers pipeline. It would be straightforward to write your own training script, though you may need to do some heavy engineering to utilize your IO/CPU and GPU/TPU/... efficiently to finish training in reasonable time.
We do plan to release the ControlNet training part in the near future.
A larger issue is how you deal with the camera pose. Because Zero123++ is all about reusing as much SD prior as possible for 3D; there can be only very weak prior for poses in SD.
Though not official, you can try the finetune code from https://github.com/TencentARC/InstantMesh
The training code is very specialized towards our infrastructure, and thus we do not have a plan to release a complete training script for now. But every module is there in the diffusers pipeline. It would be straightforward to write your own training script, though you may need to do some heavy engineering to utilize your IO/CPU and GPU/TPU/... efficiently to finish training in reasonable time.
We do plan to release the ControlNet training part in the near future.
A larger issue is how you deal with the camera pose. Because Zero123++ is all about reusing as much SD prior as possible for 3D; there can be only very weak prior for poses in SD.
Thank you for your remarkable work. I am interested in training my own ControlNet, and I'd like to use conditions beyond just depth. Could you please let me know when you might release the ControlNet training part?
