Pixelsplat

Visit Tool

pixelSplat is an open-source AI tool that generates 3D Gaussian splats from image pairs, enabling scalable and generalizable 3D reconstruction. It was recognized as a Best Paper Runner-Up at CVPR 2024.

Claim this tool

1View

At a glance

Pricing

Open Source

Free tier

Yes

API

Skill level

Technical

About

What is pixelsplat?

pixelSplat provides the code for generating 3D Gaussian splats from image pairs, a method for scalable and generalizable 3D reconstruction. Developed by David Charatan et al. and presented at CVPR 2024, this tool allows users to reconstruct 3D scenes from 2D images. The codebase has been updated to reflect the camera-ready version of the paper, including architectural improvements like per-image self-attention in the epipolar transformer, leading to slightly better results across datasets. It supports running with an arbitrary number of views, though requiring significant GPU memory for more complex setups. The project also offers pre-trained checkpoints and scripts for dataset conversion and evaluation.

Best used for

Ideal for researchers and developers who need to implement state-of-the-art 3D reconstruction from image pairs, explore novel view synthesis, and conduct experiments in computer vision. Especially valuable for academic projects and building upon foundational 3D Gaussian splatting techniques.

Common actions

reconstruct 3D scenes

generate 3D models

perform novel view synthesis

research 3D vision

workflowsdeepfakeopen-sourcecollaborationautomated workflowlow-code/no-codeface swappinggithub copilot"AI Agents"

Capabilities

Key features

3D Gaussian splatting
Image pair reconstruction
Scalable 3D reconstruction
Generalizable 3D reconstruction
Multi-GPU training support
Pre-trained checkpoints

Target Audience

professor

Integrations

Not yet documented

Pricing & Plans

Open Source

Free

FAQs

What are the hardware requirements for running pixelSplat?

Training pixelSplat, especially with the recommended configurations, requires a single GPU with 80 GB of VRAM, such as an A100 or H100. While multi-GPU training is supported, running with an arbitrary number of views may also demand substantial GPU memory.

How can I acquire the datasets used for pixelSplat?

Small subsets of the Real Estate 10k and ACID datasets, formatted for pixelSplat, are available for download. Scripts are also provided to convert your own downloaded versions of these datasets into the required format for use with the tool.

Can I use pixelSplat for novel view synthesis?

Yes, pixelSplat is designed for generalizable 3D reconstruction, which includes novel view synthesis. You can render frames from existing checkpoints using provided evaluation scripts and indices, including those for generating videos shown on the project website.

Trending

Subcategories trending in Content & Design

Image Generation AI Writing Assistants Audio & Music Video Generation Photo Editing Graphic Design

Trending

Also listed in

This tool also appears in

Research & Education › Academic Research Coding & Development › Open Source & Models

Explore

Browse AI tools by category

Content & Design Productivity & Business Coding & Development AI Agents & Automation Research & Education Wellness & Lifestyle Career Development Marketing & Growth Data & Analytics Customer Support & CX Finance E-commerce