--- license: cc-by-4.0 tags: - architecture - arrow - audio-video - contrastive - curriculum - kfold-5 - light - moderate - self-training --- # build_dataset.py ## Dataset Summary A **architecture** dataset with **audio video** modality, stored in **arrow** format. ## Preprocessing & Augmentation - **Preprocessing**: curriculum - **Augmentation**: light ## Splits & Sampling - **Split strategy**: kfold 5 - **Sampling**: contrastive ## Quality & Labeling - **Quality filtering**: moderate - **Labeling**: self training ## Files - `build_dataset.py` — main artifact of this repository ## License See the license field above.