| ---
|
| license: cc-by-4.0
|
| tags:
|
| - curriculum
|
| - huggingface
|
| - image-text
|
| - legal
|
| - mixup-cutmix
|
| - moderate
|
| - semi-auto
|
| - temporal
|
| - weighted
|
| ---
|
|
|
| # dataloader.py
|
|
|
| ## Dataset Summary
|
|
|
| A **legal** dataset with **image text** modality, stored in **huggingface** format.
|
|
|
| ## Preprocessing & Augmentation
|
|
|
| - **Preprocessing**: curriculum
|
| - **Augmentation**: mixup cutmix
|
|
|
| ## Splits & Sampling
|
|
|
| - **Split strategy**: temporal
|
| - **Sampling**: weighted
|
|
|
| ## Quality & Labeling
|
|
|
| - **Quality filtering**: moderate
|
| - **Labeling**: semi auto
|
|
|
| ## Files
|
|
|
| - `dataloader.py` — main artifact of this repository
|
|
|
| ## License
|
|
|
| See the license field above.
|
|
|
