triplane_value

TriplaneGuassian Model Card

<div align="center">

Project Page | Paper (ArXiv) | Code | Gradio demo

</div>

Introduction

TGS enables fast reconstruction from single-view image in a few seconds based on a hybrid Triplane-Gaussian 3D representation.

Examples

Results on Images Generated by Midjourney

<video controls autoplay src="https://cdn-uploads.huggingface.co/production/uploads/644dbf6453ad80c6593bf748/BcJp8alZRXAIdPmfbVGdx.qt"></video>

Results on Captured Real-world Images

<video controls autoplay src="https://cdn-uploads.huggingface.co/production/uploads/644dbf6453ad80c6593bf748/bgAxqUQpnisQAmsGZ9Q_0.qt"></video>

Model Details

The model model_lvis_rel.ckpt is trained on Objaverse-LVIS dataset, which only includes ~45K synthetic objects.

Usage

You can directly download the model in this repository or employ the model in python script by:

from huggingface_hub import hf_hub_download
MODEL_CKPT_PATH = hf_hub_download(repo_id="VAST-AI/TriplaneGaussian", filename="model_lvis_rel.ckpt", repo_type="model")

More details can be found in our Github repository.

Citation

If you find this work helpful, please consider citing our paper:

@article{zou2023triplane,
  title={Triplane Meets Gaussian Splatting: Fast and Generalizable Single-View 3D Reconstruction with Transformers},
  author={Zou, Zi-Xin and Yu, Zhipeng and Guo, Yuan-Chen and Li, Yangguang and Liang, Ding and Cao, Yan-Pei and Zhang, Song-Hai},
  journal={arXiv preprint arXiv:2312.09147},
  year={2023}
}