---
license: apache-2.0
tags:
- adaptive
- audio-video
- curriculum
- gaming
- leave-one-out
- npy-sharded
- randaugment
- self-training
- strict
---

# build_dataset.py

## Dataset Summary

A **gaming** dataset with **audio video** modality, stored in **npy sharded** format.

## Preprocessing & Augmentation

- **Preprocessing**: adaptive
- **Augmentation**: randaugment

## Splits & Sampling

- **Split strategy**: leave one out
- **Sampling**: curriculum

## Quality & Labeling

- **Quality filtering**: strict
- **Labeling**: self training

## Files

- `build_dataset.py` — main artifact of this repository

## License

See the license field above.