File size: 2,805 Bytes
a3fc62a |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 |
---
license: apache-2.0
task_categories:
- text-to-3d
- image-to-3d
language:
- en
tags:
- 4d
- 3d
- text-to-4d
- image-to-4d
- 3d-to-4d
size_categories:
- 1M<n<10M
---
# Diffusion4D: Fast Spatial-temporal Consistent 4D Generation via Video Diffusion Models
[[Project Page]](https://vita-group.github.io/Diffusion4D/) | [[Arxiv]](https://arxiv.org/abs/2405.16645) | [[Code]](https://github.com/VITA-Group/Diffusion4D)
## News
- 2024.6.28: Released rendered data from curated [objaverse-xl](https://huggingface.co/datasets/hw-liang/Diffusion4D/tree/main/objaverseXL_curated).
- 2024.6.4: Released rendered data from curated [objaverse-1.0](https://huggingface.co/datasets/hw-liang/Diffusion4D/tree/main/objaverse1.0_curated), including orbital videos of dynamic 3D, orbital videos of static 3D, and monocular videos from front view.
- 2024.5.27: Released metadata for objects!
## Overview
We collect a large-scale, high-quality dynamic 3D(4D) dataset sourced from the vast 3D data corpus of [Objaverse-1.0](https://objaverse.allenai.org/objaverse-1.0/) and [Objaverse-XL](https://github.com/allenai/objaverse-xl). We apply a series of empirical rules to filter the dataset. You can find more details in our paper. In this part, we will release the selected 4D assets, including:
1. Selected high-quality 4D object ID.
2. A render script using Blender, providing optional settings to render your personalized data.
3. Rendered 4D images by our team to save your GPU time. With 8 GPUs and a total of 16 threads, it took 5.5 days to render the curated objaverse-1.0 dataset.
## 4D Dataset ID/Metadata
We collect 365k dynamic 3D assets from Objaverse-1.0 (42k) and Objaverse-xl (323k). Then we curate a high-quality subset to train our models.
Metadata of animated objects (323k) from objaverse-xl can be found in [meta_xl_animation_tot.csv](https://huggingface.co/datasets/hw-liang/Diffusion4D/blob/main/meta_xl_animation_tot.csv).
We also release the metadata of all successfully rendered objects from objaverse-xl's Github subset in [meta_xl_tot.csv](https://huggingface.co/datasets/hw-liang/Diffusion4D/blob/main/meta_xl_tot.csv).
For text-to-4D generation, the captions are obtained from the work [Cap3D](https://huggingface.co/datasets/tiange/Cap3D).
## Citation
If you find this repository/work/dataset helpful in your research, please consider citing the paper and starring the [repo](https://github.com/VITA-Group/Diffusion4D) ⭐.
```
@article{liang2024diffusion4d,
title={Diffusion4D: Fast Spatial-temporal Consistent 4D Generation via Video Diffusion Models},
author={Liang, Hanwen and Yin, Yuyang and Xu, Dejia and Liang, Hanxue and Wang, Zhangyang and Plataniotis, Konstantinos N and Zhao, Yao and Wei, Yunchao},
journal={arXiv preprint arXiv:2405.16645},
year={2024}
}
``` |