| ---
|
| license: apache-2.0
|
| tags:
|
| - domain-specific
|
| - light
|
| - npy-sharded
|
| - semi-auto
|
| - social
|
| - stratified-90-10
|
| - strict
|
| - video-text
|
| - weighted
|
| ---
|
|
|
| # preprocess.py
|
|
|
| ## Dataset Summary
|
|
|
| A **social** dataset with **video text** modality, stored in **npy sharded** format.
|
|
|
| ## Preprocessing & Augmentation
|
|
|
| - **Preprocessing**: domain specific
|
| - **Augmentation**: light
|
|
|
| ## Splits & Sampling
|
|
|
| - **Split strategy**: stratified 90 10
|
| - **Sampling**: weighted
|
|
|
| ## Quality & Labeling
|
|
|
| - **Quality filtering**: strict
|
| - **Labeling**: semi auto
|
|
|
| ## Files
|
|
|
| - `preprocess.py` — main artifact of this repository
|
|
|
| ## License
|
|
|
| See the license field above.
|
|
|
