Hawk-Eye Innovations

[rmarcelino4]

Is there any plan to add Hawk-Eye Innovations tracking data?

[JanVanHaaren]

I'm not aware of anyone planning to add support for Hawk-Eye but a deserializer for Hawk-Eye tracking data would certainly be welcome.

[probberechts]

I have an implementation at https://github.com/probberechts/kloppy/tree/feat/hawkeye. I didn't create a PR yet because I only did some very limited testing on a single game, but it should be functional. Below is an example of how to use the API:

from pathlib import Path
from kloppy import hawkeye

hawkey_data_dir = Path(...)
season_id = 2024
league_id = 2
match_id = 1234
period_id = "*"
minute_id = "*"

# Load the 2D tracking data of the ball and the player centroids.
data_tracking = hawkeye.load(
    ball_feeds=sorted(hawkey_data_dir.glob(f'{season_id}_{league_id}_{match_id}_{period_id}_{minute_id}.football.samples.ball.gz')),
    player_centroid_feeds=sorted(hawkey_data_dir.glob(f'{season_id}_{league_id}_{match_id}_{period_id}_{minute_id}.football.samples.centroids.gz')),
    show_progress=True
)

# Loading the complete joint tracking data for a full game will probably require too much memory. 
# There are two options. 

# The first option is to load the data in batches.
# For example, to load all joint tracking data for the first minute of the game
data_joints = hawkeye.load(
    ball_feeds=[hawkey_data_dir / 'f'{season_id}_{league_id}_{match_id}_1_1.football.samples.ball.gz'],
    player_centroid_feeds=[hawkey_data_dir / f'{season_id}_{league_id}_{match_id}_1_1.football.samples.centroids.gz'],
    player_joint_feeds=[hawkey_data_dir / f'{season_id}_{league_id}_{match_id}_1_1.football.samples.joints.gz'],
    show_progress=False
)

# The second option is to load joint data only for a subset of tracking frames and or players. 
# This can be configured by providing a function.

def load_joint_data(frame, player):
    # only load joint data for a specific player
    is_player = player.jersey_no == 20
    # only load joint data for frames where the ball is closer than 2m to this player
    ball_close = frame.players_data[player].coordinates.distance_to(frame.ball_coordinates) < 2
    return is_player and ball_close
    
data_joints = hawkeye.load(
    ball_feeds=sorted(hawkey_data_dir.glob('f'{season_id}_{league_id}_{match_id}_{period_id}_{minute_id}.football.samples.ball.gz')),
    player_centroid_feeds=sorted(hawkey_data_dir.glob(f'{season_id}_{league_id}_{match_id}_{period_id}_{minute_id}.football.samples.centroids.gz')),
    player_joint_feeds=sorted(hawkey_data_dir.glob(f'{season_id}_{league_id}_{match_id}_{period_id}_{minute_id}.football.samples.joints.gz')),
    load_joint_data=load_joint_data,
    show_progress=True
)

[rmarcelino4]

Sorry for asking such a basic question, but how do I access the reference ‘from kloppy import hawkeye’?
The link doesn't seem to work and in the version of kloppy I have installed, it's not possible to import ‘hawkeye’.

[probberechts]

What do you mean by "The link doesn't seem to work"? Do you (1) get a 404 (try again, I guess; it might have been a server problem at GitHub), (2) do not have access permissions (that would be strange; I'm sure the fork is public) or (3) do not know how to install the library from a fork?

The most straightforward method to install it is through pip with:

pip install "git+https://github.com/probberechts/kloppy@feat/hawkeye"

or in editable mode (allows you to make changes to the code) with:

pip install -e "git+https://github.com/probberechts/kloppy@feat/hawkeye#egg=kloppy"