---
license: apache-2.0
task_categories:
- image-to-3d
tags:
- slam
- 3d-reconstruction
- monocular
---

This repository contains data for WildGS-SLAM: Monocular Gaussian Splatting SLAM in Dynamic Environments.

[Paper](https://huggingface.co/papers/2504.03886) | [Project Page](https://wildgs-slam.github.io/) | [Code](https://github.com/GradientSpaces/WildGS-SLAM)

WildGS-SLAM accurately tracks the camera trajectory and reconstructs a 3D Gaussian map for static elements from a monocular video sequence, effectively removing dynamic components.


### Datasets Used

WildGS-SLAM uses data from the following datasets:

* **Wild-SLAM Mocap Dataset:** ([Hugging Face](https://huggingface.co/datasets/gradient-spaces/Wild-SLAM/tree/main/Mocap))  Download instructions are available in the [github repository](https://github.com/GradientSpaces/WildGS-SLAM).
* **Wild-SLAM iPhone Dataset:** ([Hugging Face](https://huggingface.co/datasets/gradient-spaces/Wild-SLAM/tree/main/iPhone)) Download instructions are available in the [github repository](https://github.com/GradientSpaces/WildGS-SLAM).
* **Bonn Dynamic Dataset:** ([Website](https://www.ipb.uni-bonn.de/data/rgbd-dynamic-dataset/index.html)) Download instructions are available in the [github repository](https://github.com/GradientSpaces/WildGS-SLAM).
* **TUM RGB-D (dynamic) Dataset:** Download instructions are available in the [github repository](https://github.com/GradientSpaces/WildGS-SLAM).