microsoft/trocr-small-printed

1---2tags:3- trocr4- image-to-text5widget:6- src: https://layoutlm.blob.core.windows.net/trocr/dataset/SROIE2019Task2Crop/train/X00016469612_1.jpg7  example_title: Printed 18- src: https://layoutlm.blob.core.windows.net/trocr/dataset/SROIE2019Task2Crop/train/X51005255805_7.jpg9  example_title: Printed 210- src: https://layoutlm.blob.core.windows.net/trocr/dataset/SROIE2019Task2Crop/train/X51005745214_6.jpg11  example_title: Printed 312---13 14# TrOCR (small-sized model, fine-tuned on SROIE) 15 16TrOCR 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). 17 18 19## Model description20 21The 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 DeiT, while the text decoder was initialized from the weights of UniLM.22 23Images 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.24 25## Intended uses & limitations26 27You 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.28 29### How to use30 31Here is how to use this model in PyTorch:32 33```python34from transformers import TrOCRProcessor, VisionEncoderDecoderModel35from PIL import Image36import requests37 38# load image from the IAM database (actually this model is meant to be used on printed text)39url = 'https://fki.tic.heia-fr.ch/static/img/a01-122-02-00.jpg'40image = Image.open(requests.get(url, stream=True).raw).convert("RGB")41 42processor = TrOCRProcessor.from_pretrained('microsoft/trocr-small-printed')43model = VisionEncoderDecoderModel.from_pretrained('microsoft/trocr-small-printed')44pixel_values = processor(images=image, return_tensors="pt").pixel_values45 46generated_ids = model.generate(pixel_values)47generated_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]48```49 50### BibTeX entry and citation info51 52```bibtex53@misc{li2021trocr,54      title={TrOCR: Transformer-based Optical Character Recognition with Pre-trained Models}, 55      author={Minghao Li and Tengchao Lv and Lei Cui and Yijuan Lu and Dinei Florencio and Cha Zhang and Zhoujun Li and Furu Wei},56      year={2021},57      eprint={2109.10282},58      archivePrefix={arXiv},59      primaryClass={cs.CL}60}61```