README.md

---
tags:
- trocr
- image-to-text
widget:
- src: https://layoutlm.blob.core.windows.net/trocr/dataset/SROIE2019Task2Crop/train/X00016469612_1.jpg
  example_title: Printed 1
- src: https://layoutlm.blob.core.windows.net/trocr/dataset/SROIE2019Task2Crop/train/X51005255805_7.jpg
  example_title: Printed 2
- src: https://layoutlm.blob.core.windows.net/trocr/dataset/SROIE2019Task2Crop/train/X51005745214_6.jpg
  example_title: Printed 3
---
<div align="center">

# Neural Metrics · The dependable baseline for a single line of text.

<img src="https://img.shields.io/badge/Neural%20Metrics-document%20extraction-4F46E5?style=for-the-badge" alt="Neural Metrics" />
<img src="https://img.shields.io/badge/fork%20of-microsoft%2Ftrocr--base--printed-2563EB?style=flat-square" alt="fork" />

</div>

TrOCR is a transformer OCR model for cropped text lines rather than whole pages. It is old, small, fast and extremely well understood - which makes it a good control.

**We use it for:** line-level recognition after layout detection - a stable baseline that tells us whether a fancy new model is actually better.

> ### Attribution
> This is an **unmodified fork** of [`microsoft/trocr-base-printed`](https://huggingface.co/microsoft/trocr-base-printed), created by the [Qwen team](https://huggingface.co/Qwen).
> All weights, files and behaviour are identical to upstream — we rehost it so our experiments stay
> reproducible and version-pinned. The original license and all credit remain with the Qwen team.
> If you want the canonical model, please use [the original](https://huggingface.co/microsoft/trocr-base-printed).

---

<details>
<summary><b>Original model card from microsoft/trocr-base-printed</b> (click to expand)</summary>

# TrOCR (base-sized model, fine-tuned on SROIE) 

TrOCR model fine-tuned on the [SROIE dataset](https://rrc.cvc.uab.es/?ch=13). It was introduced in the paper [TrOCR: Transformer-based Optical Character Recognition with Pre-trained Models](https://arxiv.org/abs/2109.10282) by Li et al. and first released in [this repository](https://github.com/microsoft/unilm/tree/master/trocr). 

Disclaimer: The team releasing TrOCR did not write a model card for this model so this model card has been written by the Hugging Face team.

## Model description

The TrOCR model is an encoder-decoder model, consisting of an image Transformer as encoder, and a text Transformer as decoder. The image encoder was initialized from the weights of BEiT, while the text decoder was initialized from the weights of RoBERTa.

Images are presented to the model as a sequence of fixed-size patches (resolution 16x16), which are linearly embedded. One also adds absolute position embeddings before feeding the sequence to the layers of the Transformer encoder. Next, the Transformer text decoder autoregressively generates tokens.

## Intended uses & limitations

You can use the raw model for optical character recognition (OCR) on single text-line images. See the [model hub](https://huggingface.co/models?search=microsoft/trocr) to look for fine-tuned versions on a task that interests you.

### How to use

Here is how to use this model in PyTorch:

```python
from transformers import TrOCRProcessor, VisionEncoderDecoderModel
from PIL import Image
import requests

# load image from the IAM database (actually this model is meant to be used on printed text)
url = 'https://fki.tic.heia-fr.ch/static/img/a01-122-02-00.jpg'
image = Image.open(requests.get(url, stream=True).raw).convert("RGB")

processor = TrOCRProcessor.from_pretrained('microsoft/trocr-base-printed')
model = VisionEncoderDecoderModel.from_pretrained('microsoft/trocr-base-printed')
pixel_values = processor(images=image, return_tensors="pt").pixel_values

generated_ids = model.generate(pixel_values)
generated_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
```

### BibTeX entry and citation info

```bibtex
@misc{li2021trocr,
      title={TrOCR: Transformer-based Optical Character Recognition with Pre-trained Models}, 
      author={Minghao Li and Tengchao Lv and Lei Cui and Yijuan Lu and Dinei Florencio and Cha Zhang and Zhoujun Li and Furu Wei},
      year={2021},
      eprint={2109.10282},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}
```

</details>