VGGT-SLAM

Software
Navigation and mapping
Framework

VGGT-SLAM 2.0 is a dense, feed-forward SLAM system that builds 3D scene maps from images or a live RealSense camera. It targets real-time scene reconstruction.

Made by MIT-SPARK.

Source on GitHub: github.com/MIT-SPARK/VGGT-SLAM (opens in a new tab). Facts read at commit 35327ac (opens in a new tab).

Licenses

Code: BSD-2-Clause (opens in a new tab)
Read by Robotheca from LICENSE (opens in a new tab) (BSD-2-Clause) at commit 35327ac (opens in a new tab), on 22 Oct 2026.
Commercial use: Yes

What it does

Builds a map incrementally from a folder of images, visualized in Viser, with optional loop closure.
Runs online from a connected Intel RealSense camera via main_realtime.py.
Supports optional 3D open-set object detection that places bounding boxes for text queries on the map.
Includes evaluation scripts for the TUM and 7-Scenes datasets.
Builds on the VGGT model; VGGT-SLAM 1.0 remains available on the version1.0 branch.

Written by Robotheca from the project's README (opens in a new tab) at commit 35327ac.

Getting started

Difficulty: Advanced
Hardware it needs, as the project states it: A conda environment with Python 3.11 is set up by the included setup script; an Intel RealSense camera is needed only for online use.
How to start: The project's README (opens in a new tab)

Maintenance

Status: Maintained: a commit in the last 12 months
Newest commit: 29 Jun 2026 on main
Latest release: None published on GitHub

Checked on GitHub on 22 Oct 2026.

Robots that can run it

No robot on Robotheca is linked to it yet. Links come from a robot's robot.yaml, from its dependency files read as text, or from Robotheca's editors.

Robotheca lists software as a reference: nothing is hosted or run here, and every fact links to where it was read.