ocr-lab56 / README.md
srmistbiolab's picture
Upload 2 files
1ddf91b verified
|
Raw
History Blame Contribute Delete
684 Bytes
metadata
license: bsd-3-clause
tags:
  - adaptive
  - curriculum
  - education
  - image-text
  - lenient
  - light
  - npy-sharded
  - pseudo-label
  - random-90-10

dataloader.py

Dataset Summary

A education dataset with image text modality, stored in npy sharded format.

Preprocessing & Augmentation

  • Preprocessing: adaptive
  • Augmentation: light

Splits & Sampling

  • Split strategy: random 90 10
  • Sampling: curriculum

Quality & Labeling

  • Quality filtering: lenient
  • Labeling: pseudo label

Files

  • dataloader.py — main artifact of this repository

License

See the license field above.