CoWTracker

CoWTracker

Tracks every pixel of an uploaded video through time and draws the motion as colorful trajectories. Suits researchers and anyone studying how things move in footage.

Made by facebook 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: CoWTracker is a research demo that follows every pixel of a video from frame to frame. You upload a clip, pick a starting frame, and it draws the motion as colorful paths. The method is dense point tracking based on warping, and it skips the cost volume that many trackers build. It comes from researchers at Oxford's Visual Geometry Group and Meta AI.

Who it suits: It suits computer vision researchers and students, and anyone who wants to see how motion in a clip can be measured down to single pixels.

Industry | Who uses it | How they use it | Example
Computer vision | Researchers | Try dense point tracking on their own clips | Track a short clip and inspect the trajectories
Video and film | Editors and visual effects artists | See motion paths for every part of a shot | Follow a moving object through a handheld clip
Sports analysis | Analysts | Look at how bodies and objects move between frames | Track a short training clip forwards and backwards
Education | Students | Learn what motion tracking produces | Compare paths from different starting frames

Skip it if you need a commercial tool, since the license listed here limits use to non-commercial work.

Where to use it:

• A browser: open the Space's page on Hugging Face, with nothing to install
• A copy on Windows, macOS or Linux, using Python
• A Hugging Face Space with GPU hardware, which the online copy uses

Use case | Who | What they get back
Tracking a clip | Researchers | Colorful trajectories drawn over the video
Two-way tracking | Analysts | Paths that run forwards and backwards from a chosen frame
Trying settings | Students | Different looks of the same motion data

How to set it up: The online copy needs nothing. For a copy on your own computer, you need the following.

• Python with pip on Windows, macOS or Linux
• The Hugging Face hub client, which the first step installs
• The packages listed in requirements.txt
• The system packages in packages.txt, which Hugging Face installs with apt-get

1. Install the Hugging Face hub client.
2. Download the Space's files into a local folder.
3. Move into that folder.
4. Install the Gradio version the Space pins.
5. Install the remaining packages from requirements.txt.
6. Start app.py.

pip install -U huggingface_hub
hf download facebook/cowtracker --repo-type space --local-dir cowtracker
cd cowtracker
pip install gradio==6.2.0
pip install -r requirements.txt
python app.py

How to use it: Add your own video, or choose one of the examples, and press "Process Video". Then move the slider to the frame you want to track from and press "Start Tracking". Last, change the visualization settings to taste.

Tool | What it does
Process Video | Readies the uploaded clip before you pick a frame
Query frame slider | Chooses the frame the tracking starts from
Start Tracking | Follows every pixel forwards and backwards from that frame
Visualization settings | Change how the trajectories look

Example: You upload a short clip of a person walking and press "Process Video". You drag the slider to the first frame and press "Start Tracking". The Space returns the clip with colorful trajectories showing where each pixel moves.

For newcomers and experts: If you are new, start with one of the built-in example clips so you can see a result at once. Then upload a short clip of your own, since long videos take more time on shared hardware.

If you know the field, the model is a warping-based dense tracker that avoids a cost volume, and the page credits that design for its speed. You can choose any query frame and track in both time directions. The Space runs on zero-a10g hardware.

You upload a video
The clip is processed
You pick a query frame
Tracking runs in both directions
You see colorful trajectories

Good to know: The license limits who can use it, so read it before you plan any work around it.

• License: cc-by-nc-4, a Creative Commons license for non-commercial use
• Cost to run: no API key or paid service is named, and the online copy runs on zero-a10g hardware at Hugging Face
• Data it reaches: the videos you upload, which the online Space handles on Hugging Face's servers, and a local copy downloads the Space's files
• Not for: commercial use under this license, and the source lists no other exclusions

On Hugging Face, 13 November 2026

• Likes: 4
• Built with: gradio
• License: cc-by-nc-4.0

Links

• Space: https://huggingface.co/spaces/facebook/cowtracker
• Files: https://huggingface.co/spaces/facebook/cowtracker/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.