DRS-VPT

DRS-VPT

Finds where a camera image was taken inside a 3-D point cloud, without a starting guess for the pose. Meant for researchers and students who work on camera localization in 3-D scans.

Made by fulkast on Hugging Face. These notes are Mindket's, written from its Space page and README; the app's own files and license are on Hugging Face.

About

What it is: This demo finds where a camera image was taken inside a 3-D point cloud, without a starting guess for the pose. You give it an image and a point cloud, or choose an example, and it returns the camera pose in that cloud. It runs a method called DRS-VPT, which comes from a research paper.

Who it suits: It suits researchers and students who work on camera localization in 3-D scans, and anyone who wants to see the method run before reading the paper or the code. The bundled examples let you try it without preparing any data.

Industry | Who uses it | How they use it | Example
Research | Computer vision researchers | Try the method on an image and a point cloud | Check where one image sits inside a scan
Mapping | People who keep 3-D scans of places | Locate a new image within an existing point cloud | Place a recent image in the scan of the same site
Education | Students of 3-D vision | Run the demo on the bundled examples | See what a camera pose in a cloud looks like
Software | Developers who study localization | Read app.py and the examples folder | Learn how a research model is wrapped in a small demo

If you need commercial use, skip the model weights, which allow non-commercial use only.

Where to use it: The demo runs in two places.

• A web browser: open the Space page on Hugging Face, where it runs on ZeroGPU with nothing to install.
• Python 3.12 on Windows, macOS or Linux: start your own copy of the demo there.

Use case | Who | What they get back
Localize an image | Researchers | The camera pose of an image inside a point cloud
Try the examples | Students | A worked run on the sample inputs, with no data of their own
Inspect the demo | Developers | The app.py file and the examples folder to read and adapt, with the demo code under Apache 2.0

How to set it up: The online Space needs none of this. For a local copy you need a few things first.

• Python 3.12
• pip, to install packages
• A terminal on Windows, macOS or Linux

These are Hugging Face's steps for a local copy.

1. Install the Hugging Face hub package with pip.
2. Download the Space's files into a folder named drs-vpt-demo.
3. Move into that folder.
4. Install Gradio 6.30.0, then the packages listed in requirements.txt.
5. Start the app.

pip install -U huggingface_hub
hf download fulkast/drs-vpt-demo --repo-type space --local-dir drs-vpt-demo
cd drs-vpt-demo
pip install gradio==6.30.0
pip install -r requirements.txt
python app.py

How to use it: Open the Space page, or start your local copy with the command below. Give the page one image and one point cloud, or choose a bundled example, and it returns the camera pose within that cloud.

python app.py

Tool | What it does
python app.py | Starts the demo from your downloaded copy

Example: You have an image taken by a camera, together with a point cloud of the same place. You give both to the demo and it works out the pose, meaning where the camera stood and which way it faced inside the cloud. With no data at hand, you pick an example and see the same thing.

For newcomers and experts: A point cloud is a set of 3-D points that outlines a place, and a camera pose says where a camera was and where it pointed. Start by choosing one of the bundled examples and reading the pose the demo returns.

The method is published at ECCV 2026 and needs no prior pose to start from. The Space runs on ZeroGPU, reads its samples from the fulkast/drs-vpt-demo-data dataset, and loads its weights from the fulkast/drs-vpt model repository. The source does not say whether a local copy needs a GPU, so check that before you plan a local run.

You give the demo an image and a point cloud
The DRS-VPT model runs on ZeroGPU
It works out the camera pose in the cloud
The pose comes back on the demo page

Good to know: A few facts are worth checking first.

• Python: the Space and the local steps use Python 3.12.
• License: Apache 2.0 covers the demo code. The weights carry CC BY-NC 4.0, which rules out commercial use.
• Cost to run: none named. The source lists no API key, account or paid service.
• Data it reaches: it takes the image and point cloud you give it, and reads sample data and weights from two Hugging Face repositories. The source names no other account or service.
• Not for: commercial use of the model weights.

On Hugging Face, 13 November 2026

• Likes: 1
• Built with: gradio
• License: apache-2.0

Links

• Space: https://huggingface.co/spaces/fulkast/drs-vpt-demo
• Files: https://huggingface.co/spaces/fulkast/drs-vpt-demo/tree/main

Try it

Open the Space in your browser, or copy it to your own Hugging Face account from its page. Nothing in this kit runs by itself.