hebing-sjtu

hebing-sjtu / SurfSplat

Public

[2026'ICLR] Official Code for SurfSplat

46
2
100% credibility
Found Feb 04, 2026 at 22 stars 2x -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

This GitHub repository serves as a placeholder for the SurfSplat research project, which introduces a novel method for generating smooth and precise 3D reconstructions from images, with code and data scheduled for release in 2026.

How It Works

1
🔍 Discover SurfSplat

You come across SurfSplat while searching for new ways to turn everyday photos into smooth, realistic 3D scenes.

2
📖 Read the introduction

The welcome page shares an exciting research breakthrough that makes 3D models look more natural and detailed than before.

3
📄 Explore the research paper

You check out the free paper and project website, amazed by examples of flawless 3D surfaces created from simple images.

4
🌐 Watch demos

On the project site, you see videos of beautiful 3D reconstructions that feel incredibly lifelike and smooth.

5
⏳ Stay tuned for release

The team is preparing easy-to-use files and examples, promising to share everything by early 2026.

6
⭐ Follow for updates

You star the page or follow the creators to be notified the moment it's ready to try.

🎉 Create your 3D worlds

With the released tools, you easily build your own stunning 3D scenes from photos, feeling like a pro artist.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 22 to 46 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is SurfSplat?

SurfSplat is a feedforward framework using 2D Gaussian Splatting primitives to reconstruct 3D scenes with sharp geometry and detailed textures. It tackles limitations in traditional Gaussian Splatting by enforcing surface continuity priors and alpha blending, delivering anisotropic primitives for precise, coherent surfaces. This GitHub repo hosts the official code for the ICLR 2026 paper (arXiv 2602.02000), though it's currently in prep mode with no runnable code yet.

Why is it gaining traction?

With 33 stars already, it's drawing eyes as an early ICLR 2026 contender on GitHub, complete with project page and preprint. Developers chase its promise of feedforward speed over iterative splatting methods, plus superior geometry without post-processing hacks. Ties to ICLR 2026 reviewer buzz and openreview discussions amplify the hype around this official release.

Who should use this?

Computer vision researchers optimizing 3D reconstruction pipelines for novel view synthesis. Teams building real-time rendering apps needing high-fidelity surfaces from scans, like AR/VR devs ditching slow NeRFs. ScanNet dataset users eyeing preprocessed data for benchmarks.

Verdict

Skip for now—1.0% credibility score reflects the empty repo (just README) and code drop delayed to Feb 2026. Stars and docs are nascent, but track for ICLR 2026 openreview github leaks or prediction lookups if Gaussian Splatting is your jam; it'll mature fast post-release.

(178 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.