We introduce Sparc3D, a unified framework that combines a sparse deformable marching cubes representation Sparcubes with a novel encoder Sparconv-VAE. Sparcubes converts raw meshes into high-resolution (1024³) surfaces with arbitrary topology by scattering signed distance and deformation fields onto a sparse cube, allowing differentiable optimization. Sparconv-VAE is the first modality-consistent variational autoencoder built entirely upon sparse convolutional networks, enabling efficient and near-lossless 3D reconstruction suitable for high-resolution generative modeling through latent diffusion.

If you find this work useful, please cite:

@article{li2025sparc3d,
  title   = {Sparc3D: Sparse Representation and Construction for High-Resolution 3D Shapes Modeling},
  author  = {Li, Zhihao and Wang, Yufei and Zheng, Heliang and Luo, Yihao and Wen, Bihan},
  journal = {arXiv preprint arXiv:2505.14521},
  year    = {2025}
}