| ---
|
| license: bsd-3-clause
|
| tags:
|
| - adaptive
|
| - contrastive
|
| - domain-specific
|
| - kfold-5
|
| - lmdb
|
| - mixup-cutmix
|
| - movie-posters
|
| - pointcloud-text
|
| - semi-auto
|
| ---
|
|
|
| # preprocess.py
|
|
|
| ## Dataset Summary
|
|
|
| A **movie posters** dataset with **pointcloud text** modality, stored in **lmdb** format.
|
|
|
| ## Preprocessing & Augmentation
|
|
|
| - **Preprocessing**: domain specific
|
| - **Augmentation**: mixup cutmix
|
|
|
| ## Splits & Sampling
|
|
|
| - **Split strategy**: kfold 5
|
| - **Sampling**: contrastive
|
|
|
| ## Quality & Labeling
|
|
|
| - **Quality filtering**: adaptive
|
| - **Labeling**: semi auto
|
|
|
| ## Files
|
|
|
| - `preprocess.py` — main artifact of this repository
|
|
|
| ## License
|
|
|
| See the license field above.
|
|
|
