---
license: bsd-3-clause
tags:
- aggressive
- audio-text
- comics
- contrastive
- light
- moderate
- npy-sharded
- semi-auto
- temporal
---

# build_dataset.py

## Dataset Summary

A **comics** dataset with **audio text** modality, stored in **npy sharded** format.

## Preprocessing & Augmentation

- **Preprocessing**: aggressive
- **Augmentation**: light

## Splits & Sampling

- **Split strategy**: temporal
- **Sampling**: contrastive

## Quality & Labeling

- **Quality filtering**: moderate
- **Labeling**: semi auto

## Files

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

## License

See the license field above.