Awesome

SoccerNet Ball Action Spotting Challenge 2024 Baseline

This repo is a fork of the 1st place solution for the SoccerNet Ball Action Spotting Challenge 2023.

The 2024 Ball Action Spotting baseline was trained using weights from stage 2 of the solution above and transfer learning approach as desribed in step 3, with changes in focal loss with parameters gamma 1.0 and alpha 0.5.

Thanks to the amazing work of the 1st place winner og the 2023 Ball Action Spotting Challenge Ruslan Baikulov!

Quick setup and start

Requirements

• Linux (tested on Ubuntu 20.04)
• NVIDIA GPU (pipeline tuned for RTX 3080)
• NVIDIA Drivers >= 520, CUDA >= 11.8
• Docker
• NVIDIA Container Toolkit

Run

Clone the repo and enter the folder.

git clone git@github.com:recokick/ball-action-spotting.git
cd ball-action-spotting

Build a Docker image and run a container.

make  # stop, build, run

# do the same
make stop
make build
make run

make  # by default GPU 0 passed
make GPUS=all  # all GPUs
make GPUS=none  # without GPUs

make run GPUS=2  # pass the first two GPUs
make run GPUS='\"device=1,2\"'  # pass GPUs numbered 1 and 2

make logs
make exec  # run a new command in a running container
make exec COMMAND="bash"  # do the same
make stop

make

From now on, you should run all commands inside the docker container.

Download sampling_weights_001 and action_sampling_weights_002 from author's Google Drive and copy the files to the data directory so that the folder structure is as follows:

data
├── action
│   ├── experiments
│   │   └── action_sampling_weights_002
│   └── predictions
│       └── action_sampling_weights_002
├── ball_action
│   ├── experiments
│   │   └── sampling_weights_001
│   └── predictions
│       └── sampling_weights_001
├── readme_images
└── soccernet
    └── spotting-ball-2024
        └── england_efl

OR

Download the Ball Action Spotting 2023 dataset and Action Spotting 2023 dataset if you want to train the models from scratch. To get the password, you must fill NDA (link).

Now you can train models and use them to predict games. To reproduce the final solution, you can use the following commands (for the --experiment sampling_weights_001 parts of the steps you might want to change the constant soccernet_dir in src/ball_action/constants.py to soccernet_dir / "spotting-ball-2023"):

# Train and predict basic experiment on all folds
python scripts/ball_action/train.py --experiment sampling_weights_001
python scripts/ball_action/predict.py --experiment sampling_weights_001

# Training on Action Spotting Challenge dataset
python scripts/action/train.py --experiment action_sampling_weights_002

# Transfer learning
python scripts/ball_action/train.py --experiment ball_tuning_001
python scripts/ball_action/predict.py --experiment ball_tuning_001
python scripts/ball_action/evaluate.py --experiment ball_tuning_001
python scripts/ball_action/predict.py --experiment ball_tuning_001 --challenge
python scripts/ball_action/ensemble.py --experiments ball_tuning_001 --challenge
# To train models without the use of the test subset of the data in training use argument --folds train 

# Spotting results will be there
cd data/ball_action/predictions/ball_tuning_001/challenge/ensemble/
zip results_spotting.zip ./*/*/*/results_spotting.json