LaSoDa

 

Research  

 

Labeled Soccer Database (LaSoDa)

 

 

Description

LaSoDa is composed by 60 annotated images from matches in five stadiums with different characteristics (positions of the cameras, grass colors, and illumination setups) and light conditions (day and night).

The main characteristics of this set of images are the following:

- Their content exhaustively covers all areas of the field of play: the four corners and the central part.

- They show five different zoom levels.

- They have been acquired with four different types of cameras: master camera, side camera, end camera, and aerial camera.

futbol

For questions about this database, please contact Carlos Cuevas at This email address is being protected from spambots. You need JavaScript enabled to view it..

Field of play model

fut2

The images in LaSoDa correspond to matches played in stadiums with the field markings established by the “Laws of the Game” of the IFAB (https://www.theifab.com/). According to these rules, the model of the field of play with which the LaSoDa images are related to is the one depicted in this figure, which is characterized by 151 key points that correspond to: 

- 6 longitudinal straight lines (indicated with L).

- 21 transverse straight lines (indicated with a T).

- 3 circular lines (indicated with a C).

This model of the field of play and the coordinates of the 151 key points are included among the database download material.

 

Ground truth description

For each image in the database, a .MAT file (Matlab format) with its ground truth is included in the database, whose content is the following:

- WLM - White Line Marks:

- Structure of cells where each row provides the information of one of the white lines in the image.

- The second column contains the list of indexes indicating what pixels are part of the line.

- The first column indicates the identifier of the line, which can be one of the following. 

Captura

- GLM - Grass Line Marks:

- Structure of cells where each row provides the information of one of the lines in the image that separates the grass bands resulting from bending the grass in different directions.

- The second column contains the list of indexes indicating what pixels are part of the line.

- The first column indicates the identifier of the line which, as can be seen in the model of the field of play, can be one of the following: T2, T3, …, T10, T12, T13, …, T20.

- Hom:

- 3x3 matrix that corresponds to the homography transformation that allows transforming the original image to the canvas of the field of play model.

-FoP:

- Binary mask indicating what area of the image corresponds to the field of play.

 

Download

Click here to download the LaSoDa database (images and ground truth).

Click here to download the image of model of the field of play and the coordinates of the 151 key points in the image.

 

Disclaimer

The images in this database have been sampled from matches that are publicly available on the YouTube channel of FIFA. While the field model and ground truth are original work, the authors claim no rights over the images themselves, which remain property of their original copyright holder, and merely make them available for convenience of the academic community.

 

Citation

C. Cuevas, D. Berjón, N. García, “Grass band detection in soccer images for improved image registration”, Signal Processing: Image Communication, vol. 109, article 116837, pp. 1-16, Nov. 2022.