Skip to content
ModelListedExternal
Download

PixARMesh EdgeRunner

PixARMesh is a mesh-native autoregressive framework for single-view 3D scene reconstruction. Instead of reconstructing via intermediate volumetric or implicit representations, PixARMesh directly models instances with...

Sourcehuggingface.co/zx1239856/PixARMesh-EdgeRunner/tree/e9ea13d297a35d833d3c24e1b76ff202b0eb2c69
KindText modelBase modelzx1239856/edgerunnerVersionmodelBgrade

Model source

Kind
Text model
File size
2.86 GB
Base model
zx1239856/edgerunner
Version
vmodel
Updated
Oct 10, 2026
Downloads
67
Source
Hugging Face

Source description

PixARMesh: Autoregressive Mesh-Native Single-View Scene Reconstruction

PixARMesh is a mesh-native autoregressive framework for single-view 3D scene reconstruction. Instead of reconstructing via intermediate volumetric or implicit representations, PixARMesh directly models instances with native mesh representation. Object poses and meshes are predicted in a unified autoregressive sequence.

Model Details

Reconstructing complete 3D indoor scenes from a single RGB image is a complex task. PixARMesh jointly predicts object layout and geometry within a unified model, producing coherent and artist-ready meshes in a single forward pass. By augmenting a point-cloud encoder with pixel-aligned image features and global scene context via cross-attention, the model enables accurate spatial reasoning. Scenes are generated autoregressively from a unified token stream containing context, pose, and mesh, yielding compact meshes with high-fidelity geometry.

Citation

If you find PixARMesh useful in your research, please consider citing:

@inproceedings{zhang2026pixarmesh,
  author    = {Zhang, Xiang and Yoo, Sohyun and Wu, Hongrui and Li, Chuan and Xie, Jianwen and Tu, Zhuowen},
  title     = {PixARMesh: Autoregressive Mesh-Native Single-View Scene Reconstruction},
  booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  month     = {June},
  year      = {2026},
  pages     = {5881-5891}
}

Acknowledgements

PixARMesh builds upon several excellent open-source projects including Grounded-Segment-Anything, Depth Pro, DINOv2, and weights from EdgeRunner and BPT. We also use physically-based renderings from the 3D-FRONT scenes.

Sources

1 source
Hugging FacePrimaryVersion e9ea13d297a35d833d3c24e1b76ff202b0eb2c69
huggingface.co/zx1239856/PixARMesh-EdgeRunner/tree/e9ea13d297a35d833d3c24e1b76ff202b0eb2c69

Verified Oct 10

Model artifacts

1 artifact

model.safetensors

safetensors · 2.86 GB · SHA-256 d63132aefdc3…e424 · Hugging Face

Download

Source excerpts

3 excerpts
ReadmeAug 1huggingface.co/zx1239856/PixARMesh-EdgeRunner

--- base_model: - zx1239856/edgerunner datasets: - zx1239856/3d-front-ar-packed library_name: transformers license: cc-by-sa-4.0 pipeline_tag: image-to-3d --- # PixARMesh: Autoregressive Mesh-Native Single-View Scene Reconstruction PixARMesh is a mesh-native autoregressive framework for single-view 3D scene reconstruction. Instead of reconstructing via intermediate volumetric or implicit representations, PixARMesh directly models instances with native mesh representation. Object poses and meshes are predicted in a unified autoregressive sequence. ## Model Details Reconstructing complete 3D indoor scenes from a single RGB image is a complex task. PixARMesh jointly predicts object layout and geometry within a unified model, producing coherent and artist-ready meshes in a single forward pass. By augmenting a point-cloud encoder with pixel-aligned image features and global scene context via cross-attention, the model enables accurate spatial reasoning. Scenes are generated autoregressively from a unified token stream containing context, pose, and mesh, yielding compact meshes with high-fidelity geometry. - **Repository:** https://github.com/mlpc-ucsd/PixARMesh - **Paper:** [arXiv:2603.05888](https://arxiv.org/abs/2603.05888) - **Project Page:** https://mlpc-ucsd.github.io/PixARMesh/ ## Citation If you find PixARMesh useful in your research, please consider citing: ```bibtex @inproceedings{zhang2026pixarmesh, author = {Zhang, Xiang and Yoo, Sohyun and Wu, Hongrui and Li, Chuan and Xie, Jianwen and Tu, Zhuowen}, title = {PixARMesh: Autoregressive Mesh-Native Single-View Scene Reconstruction}, booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)}, month = {June}, year = {2026}, pages = {5881-5891} } ``` ## Acknowledgements PixARMesh builds upon several excellent open-source projects including [Grounded-Segment-Anything](https://github.com/IDEA-Research/Grounded-Segment-Anything), [Depth Pro](https://github.com/apple/ml-depth-pro), [DINOv2](https://github.com/facebookresearch/dinov2), and weights from [EdgeRunner](https://github.com/NVlabs/EdgeRunner) and [BPT](https://github.com/Tencent-Hunyuan/bpt). We also use physically-based renderings from the [3D-FRONT](https://tianchi.aliyun.com/specials/promotion/alibaba-3d-scene-dataset) scenes.

NoteOct 10huggingface.co/zx1239856/PixARMesh-EdgeRunner

Source context: 67 downloads · 0 likes · Pipeline image-to-3d · Library transformers · Repo zx1239856/PixARMesh-EdgeRunner

NoteAug 1huggingface.co/zx1239856/PixARMesh-EdgeRunner

Source context: 3893 downloads · 0 likes · Pipeline image-to-3d · Library transformers · Repo zx1239856/PixARMesh-EdgeRunner