unlimited-ocr-torch#

Unlimited-OCR document parser that reads a page into layout elements (titles, paragraphs, tables, figures) with recognized text, including long and multi-column documents.

Details

  • Model name: unlimited-ocr-torch

  • Model source: https://huggingface.co/baidu/Unlimited-OCR

  • Model author: Baidu

  • Model license: MIT

  • Model size: 6.21 GB

  • Exposes embeddings? no

  • Tags: ocr, detection, torch, transformer, vlm

Requirements

  • Packages: torch, torchvision, transformers>=4.49,<5, easydict, einops, addict

  • CPU support

    • yes

  • GPU support

    • yes

Example usage

 1import fiftyone as fo
 2import fiftyone.zoo as foz
 3
 4dataset = foz.load_zoo_dataset(
 5    "coco-2017",
 6    split="validation",
 7    dataset_name=fo.get_default_dataset_name(),
 8    max_samples=50,
 9    shuffle=True,
10)
11
12model = foz.load_zoo_model("unlimited-ocr-torch")
13
14dataset.apply_model(model, label_field="predictions")
15
16session = fo.launch_app(dataset)