can you provide the training details on the dna-rendering dataset?

[liwenssss]

[yutian10]

Hi! Could you clarify which specific training details you’re looking for regarding the DNA-rendering dataset?

[liwenssss]

Hi! Could you clarify which specific training details you’re looking for regarding the DNA-rendering dataset?

Hi! Thanks for your reply. I’d like to ask whether you plan to release the data preprocessing and data loading code, as well as the training scripts used on the DNA-rendering dataset.

[yutian10]

This process can be reproduced as follows:
Regarding dataset preparation, you can directly download the DNA-Rendering Dataset and use the 4DGS model provided in this repository (see https://github.com/fudan-zvg/4d-gaussian-splatting?tab=readme-ov-file) to generate clean-noisy training pairs: subsample frames for training and then render over the full timestamp range . To train the video diffusion model, refer to the LoRA training script in https://github.com/modelscope/DiffSynth-Studio/tree/main/examples/wanvideo.