OpenSportsLab/OSL-cls-action-mvitv2

⚽ OpenSportsLib Classification Model (MViT - V2)

📌 Overview

This model is a video-based classification model built using the OpenSportsLib, designed for soccer action classification.

• Task: Action / Event Classification
• Architecture: MViT (Multiscale Vision Transformer)
• Library: OpenSportsLib
• Input: Video clips

---

📂 Dataset

Training Dataset

This model is trained on the SoccerNet – MVFouls (Classification subset):

👉 https://huggingface.co/datasets/OpenSportsLab/soccernetpro-classification-vars/tree/mvfouls/train

• Domain: Soccer video understanding
• Task: Event classification
• Modality: Video

---

📊 Benchmark Results

Metric	Score
Accuracy	0.57
Balanced Accuracy	0.4
Top-2	0.78

---

🔧 Using with OpenSportsLib

For more details about OpenSportsLib visit the below link

👉 Github - https://github.com/OpenSportsLab/opensportslib

👉 PyPi - https://pypi.org/project/opensportslib/

👉 Documentations - https://opensportslab.github.io/opensportslib/

Import the library

import opensportslib
print("OpenSportsLib imported successfully")

Run inference

from opensportslib.apis import ClassificationModel

my_model = ClassificationModel(
    config="/path/to/classification.yaml",
👉   weights="OpenSportsLab/OSL-cls-action-mvitv2",
)

predictions = my_model.infer(
    test_set="/path/to/test.json",
)

saved_predictions = my_model.save_predictions(
    output_path="/path/to/predictions.json",
    predictions=predictions,
)

metrics = my_model.evaluate(
    test_set="/path/to/test.json",
    predictions=saved_predictions,
)

print(metrics)

---

📜 License

• Open source license: AGPL 3.0 for research, academic, and community use.

• Commercial license: For proprietary or commercial deployment, please contact the project maintainers.

__

📎 Citation

@misc{opensportslib_mvitv2_classification,
  title={OpenSportsLib Classification MViT V2},
  author={OpenSportsLab},
  year={2026},
  howpublished={https://huggingface.co/OpenSportsLab/oslib-MViTv2-classification}
}

---

🙏 Acknowledgements

• Dataset: SoccerNet / OpenSportsLab
• Library: https://github.com/OpenSportsLab/opensportslib