HockeyRink

HockeyRink

Finds 56 landmark points on an ice hockey rink in a picture and draws them with IDs and confidence scores. Built for sports analysts, broadcast graphics teams and vision researchers.

Made by SimulaMet-HOST 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 Space finds 56 landmark points on an ice hockey rink in a picture and draws them on the image, each with an ID and a confidence score. It runs a YOLOv8 pose model trained on a rink dataset. That saves you from marking rink lines by hand before any spatial analysis.

Who it suits: It suits sports analysts, broadcast graphics teams and computer vision researchers who need to know where the rink sits in a camera frame.

Title: Who might use it
Industry | Who uses it | How they use it | Example
Sports broadcasting | Graphics teams | Anchor overlays to the rink surface | Placing a graphic on the ice from a wide camera shot
Sports analytics | Analysts | Map players onto a rink plan | Turning a frame into rink coordinates for game statistics
Computer vision | Researchers | Test keypoint detection on sports footage | Checking how it copes when players block landmarks
AR and VR | Developers | Line up virtual scenes with a real rink | Fitting a virtual scene to a rink view

People who work with other sports, or with no rink images at all, can skip it.

Where to use it:

• A web browser: open the Space page on Hugging Face and upload an image
• Hugging Face Spaces: it runs there on a basic CPU machine
• Windows, macOS or Linux: run your own copy with Python

Title: Jobs it does
Use case | Who | What they get back
Camera calibration | Vision engineers | Rink landmark points to feed homography estimation
Broadcast overlays | Graphics teams | Keypoint positions that show where the rink lies in the frame
Player tracking | Analysts | Fixed rink reference points to place players against

How to set it up: Using the online Space needs nothing installed. To run a copy on your own computer you need these first:

• Python on Windows, macOS or Linux
• The packages in requirements.txt, with Gradio 5.12.0

1. Install the Hugging Face hub package.
2. Download the Space into a folder named HockeyRink.
3. Move into that folder.
4. Install Gradio 5.12.0, then the other packages.
5. Start the app.

pip install -U huggingface_hub
hf download SimulaMet-HOST/HockeyRink --repo-type space --local-dir HockeyRink
cd HockeyRink
pip install gradio==5.12.0
pip install -r requirements.txt
python app.py

How to use it: Online, open the Space page and upload a picture of a hockey rink. On your own computer, start the app with this command and use it the same way.

python app.py

The model detects the keypoints and shows each one on the picture with its ID and confidence score. You can feed those results into later spatial work or plots.

Example: You upload a wide-angle frame of a game where players partly block the rink. The Space returns the same picture with the detected keypoints marked, each labeled with its ID and confidence score.

For newcomers and experts: If you are new to this, think of it as a tool that finds fixed rink landmarks in a photo. Open the Space page and try one of its example pictures, or upload your own rink image.

If you work in the field, the model is a YOLOv8-Large pose estimator that takes RGB images at any resolution and returns 56 keypoint coordinates, each paired with a confidence score. The card reports average mAP at 0.5 of 97.48%, mAP from 0.5 to 0.95 of 76.45%, precision of 96.21% and recall of 96.24%. It was trained on Swedish Hockey League footage, and the card gives no results for rinks from other leagues.

Upload a rink image
The model finds 56 keypoints
Each point gets an ID and a score
The app draws them on the image

Good to know: The card lists speeds measured on two machines.

Title: Speed in frames per second
Tesla T4 GPU: 13.64 FPS
M3 MacBook Pro: 6.4 FPS

• License: MIT
• Cost to run: none named
• Data it reaches: the images you upload to the Space or to your own copy

On Hugging Face, 13 November 2026

• Likes: 1
• Built with: gradio
• License: mit

Links

• Space: https://huggingface.co/spaces/SimulaMet-HOST/HockeyRink
• Files: https://huggingface.co/spaces/SimulaMet-HOST/HockeyRink/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.