liminghao1630 commited on
Commit
31a5389
1 Parent(s): 244f438

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +53 -0
README.md ADDED
@@ -0,0 +1,53 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ tags:
3
+ - trocr
4
+ - image-to-text
5
+ ---
6
+
7
+ # TrOCR (large-sized model, fine-tuned on STR benchmarks)
8
+
9
+ TrOCR model fine-tuned on the training sets of IC13, IC15, IIIT5K, SVT. 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).
10
+
11
+ ## Model description
12
+
13
+ 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.
14
+
15
+ 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.
16
+
17
+ ## Intended uses & limitations
18
+
19
+ 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.
20
+
21
+ ### How to use
22
+
23
+ Here is how to use this model in PyTorch:
24
+
25
+ ```python
26
+ from transformers import TrOCRProcessor, VisionEncoderDecoderModel
27
+ from PIL import Image
28
+ import requests
29
+
30
+ # load image from the IIIT-5k dataset
31
+ url = 'https://i.postimg.cc/ZKwLg2Gw/367-14.png'
32
+ image = Image.open(requests.get(url, stream=True).raw).convert("RGB")
33
+
34
+ processor = TrOCRProcessor.from_pretrained('microsoft/trocr-large-str')
35
+ model = VisionEncoderDecoderModel.from_pretrained('microsoft/trocr-large-str')
36
+ pixel_values = processor(images=image, return_tensors="pt").pixel_values
37
+
38
+ generated_ids = model.generate(pixel_values)
39
+ generated_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
40
+ ```
41
+
42
+ ### BibTeX entry and citation info
43
+
44
+ ```bibtex
45
+ @misc{li2021trocr,
46
+ title={TrOCR: Transformer-based Optical Character Recognition with Pre-trained Models},
47
+ author={Minghao Li and Tengchao Lv and Lei Cui and Yijuan Lu and Dinei Florencio and Cha Zhang and Zhoujun Li and Furu Wei},
48
+ year={2021},
49
+ eprint={2109.10282},
50
+ archivePrefix={arXiv},
51
+ primaryClass={cs.CL}
52
+ }
53
+ ```