Questions about training data format and recommended GPU specifications

[demian-j-lim]

Hi, thanks for your great work on EasyOCR!
I am planning to train a custom OCR model using EasyOCR, and I have a few questions:

Data Format
What is the recommended data format when preparing datasets for training?
(e.g., directory structure, annotation format like JSON, CSV, or specific labeling requirements)

GPU Requirements

What is the minimum VRAM specification needed for training a small to medium-sized model?

What would be the recommended VRAM and GPU spec for comfortable training (for example, for large datasets)?

If there are any official references, documentation, or example datasets for training, it would be very helpful if you could share them.

Thank you in advance!