OCR0003: Benchmark data for Woodblock printed OCR images

[tenzinyonten]

Description:

Upload the Derge dataset images to AWS S3, generate metadata CSVs, and upload the CSVs to a Hugging Face repository.

Implementation:

1. Upload Images: Upload images to the S3 bucket and generate public URLs.
2. Create CSVs: Generate train, test, and eval CSVs with metadata.
3. Upload to Hugging Face: Upload the CSVs to the Hugging Face repository.

Subtasks:

• Upload images to S3.
• Create metadata CSVs (with filenames, URLs and transcripts).
• Convert CSV files to Parquet format and upload them to the Hugging Face repository.

Completion Criteria:

• Images uploaded to AWS S3 with URLs.
• Metadata CSVs created and uploaded to Hugging Face.

Card Reviewer:

• @ta4tsering

[tenzinyonten]

Dergetenjur data
https://huggingface.co/datasets/openpecha/OCR-Dergetenjur

[ta4tsering]

please out all the images in two or three different folders name it derge_tejur_01, 02 and so on. Zip them and upload to the hugging face repo. just like in the other repos https://huggingface.co/datasets/openpecha/OCR-KhyentseWangpo/tree/main