martinjolif/football-pitch-detection

Football Pitch Keypoint Detection Dataset

Dataset Description

This dataset consists of images captured from a football match in a stadium, taken from a camera positioned near the middle of the pitch. Each image contains annotations for 32 pitch keypoints:

<img src="pitch_keypoints.png" alt="image" width="600">

For every keypoint in an image, a position is provided using the standard YOLO format:

• —x: X-coordinate of the keypoint position
• —y: Y-coordinate of the keypoint position
• — This dataset is suitable for tasks such as pose estimation, and can be used to detect football pitch keypoints in images or videos of football games.

Usage

Download dataset in your current working directory `cwd`:

hf download martinjolif/football-ball-detection --repo-type dataset --local-dir cwd

Finetune on this dataset:

yolo pose train data=data/data.yaml model=yolo11n.pt epochs=1 batch=32 imgsz=640 device=mps

Validate custom-trained model:

yolo pose val model=path/to/best.pt

References

The dataset comes from Roboflow Universe, where you can also visualize the position of the keypoints and how they are connected.

Cite the dataset:

@misc{
  football-field-detection-f07vi-d0ele_dataset,
  title = { football-field-detection Dataset },
  type = { Open Source Dataset },
  author = { Football project },
  howpublished = { \url{ https://universe.roboflow.com/football-project-pifbc/football-field-detection-f07vi-d0ele } },
  url = { https://universe.roboflow.com/football-project-pifbc/football-field-detection-f07vi-d0ele },
  journal = { Roboflow Universe },
  publisher = { Roboflow },
  year = { 2025 },
  month = { oct },
  note = { visited on 2025-12-12 },
  }