Minghao Li commited on
Commit
c92d791
β€’
1 Parent(s): f31d49e
Files changed (5) hide show
  1. app.py +45 -0
  2. image_0.png +0 -0
  3. image_1.png +0 -0
  4. image_2.png +0 -0
  5. requirements.txt +4 -0
app.py ADDED
@@ -0,0 +1,45 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # import os
2
+ # os.system('pip install git+https://github.com/huggingface/transformers.git --upgrade')
3
+
4
+ import gradio as gr
5
+ from transformers import TrOCRProcessor, VisionEncoderDecoderModel
6
+ import requests
7
+ from PIL import Image
8
+
9
+ processor = TrOCRProcessor.from_pretrained("microsoft/trocr-base-printed")
10
+ model = VisionEncoderDecoderModel.from_pretrained("microsoft/trocr-base-printed")
11
+ model.config.eos_token_id = 2
12
+
13
+ # load image examples
14
+ urls = ['https://layoutlm.blob.core.windows.net/trocr/dataset/SROIE2019Task2Crop/train/X00016469612_1.jpg', 'https://layoutlm.blob.core.windows.net/trocr/dataset/SROIE2019Task2Crop/train/X51005255805_7.jpg', 'https://layoutlm.blob.core.windows.net/trocr/dataset/SROIE2019Task2Crop/train/X51005745214_6.jpg']
15
+ for idx, url in enumerate(urls):
16
+ image = Image.open(requests.get(url, stream=True).raw)
17
+ image.save(f"image_{idx}.png")
18
+
19
+ def process_image(image):
20
+ # prepare image
21
+ pixel_values = processor(image, return_tensors="pt").pixel_values
22
+
23
+ # generate (no beam search)
24
+ generated_ids = model.generate(pixel_values)
25
+
26
+ # decode
27
+ generated_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
28
+
29
+ return generated_text
30
+
31
+ title = "Interactive demo: TrOCR"
32
+ description = "Demo for Microsoft's TrOCR, an encoder-decoder model consisting of an image Transformer encoder and a text Transformer decoder for state-of-the-art optical character recognition (OCR) on single-text line images. This particular model is fine-tuned on SROIE Task 2, a dataset of annotated printed images. To use it, simply upload a (single-text line) image or use one of the example images below and click 'submit'. Results will show up in a few seconds."
33
+ article = "<p style='text-align: center'><a href='https://arxiv.org/abs/2109.10282'>TrOCR: Transformer-based Optical Character Recognition with Pre-trained Models</a> | <a href='https://github.com/microsoft/unilm/tree/master/trocr'>Github Repo</a></p>"
34
+ examples =[["image_0.png"], ["image_1.png"], ["image_2.png"]]
35
+
36
+ #css = """.output_image, .input_image {height: 600px !important}"""
37
+
38
+ iface = gr.Interface(fn=process_image,
39
+ inputs=gr.inputs.Image(type="pil"),
40
+ outputs=gr.outputs.Textbox(),
41
+ title=title,
42
+ description=description,
43
+ article=article,
44
+ examples=examples)
45
+ iface.launch(debug=True)
image_0.png ADDED
image_1.png ADDED
image_2.png ADDED
requirements.txt ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+ gradio
2
+ torch
3
+ Pillow
4
+ transformers