Mvsplat

Visit Tool

MVSplat is an open-source computer vision tool that enables efficient 3D Gaussian Splatting from sparse multi-view images. It allows for the creation of 3D models from a limited number of input views.

Claim this tool

No Views Yet

At a glance

Pricing

Open Source

Free tier

Yes

API

Skill level

Technical

About

What is mvsplat?

MVSplat is an open-source project presented at ECCV 2024, designed for efficient 3D Gaussian Splatting from sparse multi-view images. It provides a robust framework for generating 3D models from a limited set of input images, making it valuable for computer vision research and 3D graphics development. The tool supports evaluation, video rendering, and training with various datasets like RealEstate10K, ACID, and DTU. It also offers features for fine-tuning from released weights and conducting ablations to understand model components. MVSplat is built on PyTorch Lightning, allowing for scalable training on multiple GPUs and nodes, and has extensions like PanSplat for higher resolutions and MVSplat360 for 360° NVS from fewer views.

Best used for

Ideal for developers and data scientists who need to efficiently generate 3D models from sparse multi-view images, render novel views, and train advanced computer vision models. Especially valuable for research in 3D reconstruction and novel view synthesis, offering scalable training capabilities.

Common actions

create 3D models

render 3D scenes

train computer vision models

open-sourcelow-code/no-codedeepfakeautomated workflowworkflowscollaborationface swapping"AI Agents"github copilot

Capabilities

Key features

3D Gaussian Splatting
Sparse multi-view input
Novel view synthesis
Video rendering
Scalable training
Cross-dataset generalization

Target Audience

developerdata scientist

Integrations

Not yet documented

Pricing & Plans

Open Source

Free

FAQs

What are the hardware requirements for training MVSplat models?

MVSplat models are trained with a single A100 (80GB) GPU, but can also be trained on multiple GPUs with smaller RAM by adjusting the batch size. It supports multi-node training on SLURM clusters.

How can I acquire the datasets used for MVSplat training?

MVSplat uses RealEstate10K and ACID datasets, similar to pixelSplat. Small subsets are available for download, and scripts are provided to convert downloaded versions to the required format. Preprocessed versions are also linked.

Can MVSplat be used for 360-degree novel view synthesis?

Yes, an upgraded version called MVSplat360, presented at NeurIPS '24, combines video diffusion to achieve 360° novel view synthesis for large-scale scenes from as few as 5 input views.

Trending

Subcategories trending in Content & Design

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

Trending

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