#SparConV
*Here's the paper, and there's slopject job #1, apparently -- a megaton of Chinese sci-fi toys.

*Well, this oughta keep me busy a while. This is gonna be a hot summer #techart #curator #designtheory #realization #SparConV

arxiv.org/pdf/2505.145...
July 3, 2025 at 3:44 AM
Turning messy 3D meshes into crisp, high-res magic—meet Sparc3D 🧠📦

www.threads.com/@luokai/post...
luokai (@luokai) on Threads
Turning messy 3D meshes into crisp, high-res magic—meet Sparc3D 🧠📦 Sparc3D ditches bulky voxel grids & lossy VAEs with Sparcubes and Sparconv-VAE—a sparse pipeline for efficient, detail-rich 3D shape...
www.threads.com
June 23, 2025 at 3:42 PM
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 [4/6 of https://arxiv.org/abs/2505.14521v1]
May 21, 2025 at 6:09 AM
introduce SparC, a unified framework that combines a sparse deformable marching cubes representation SparseCubes with a novel encoder SparConv-VAE. SparseCubes converts raw meshes into high-resolution ($1024^3$) surfaces with arbitrary topology by [3/6 of https://arxiv.org/abs/2505.14521v1]
May 21, 2025 at 6:09 AM
Sparc3D ditches bulky voxel grids & lossy VAEs with Sparcubes and Sparconv-VAE—a sparse pipeline for efficient, detail-rich 3D shape modeling. It reconstructs 1024³ surfaces with complex topology while staying fast, light, and diffusion-friendly.

Project: lizhihao6.github.io/Sparc3D/
Sparc3D: Sparse Representation and Construction for High-Resolution 3D Shapes Modeling
Sparc3D: Sparse Representation and Construction for High-Resolution 3D Shapes Modeling
lizhihao6.github.io
June 23, 2025 at 3:42 PM