merve HF staff commited on
Commit
1689fdb
1 Parent(s): ceeec68

Create app.py

Browse files
Files changed (1) hide show
  1. app.py +125 -0
app.py ADDED
@@ -0,0 +1,125 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ from huggingface_hub import hf_hub_download
2
+ import re
3
+ from PIL import Image
4
+ import requests
5
+ from nougat.dataset.rasterize import rasterize_paper
6
+
7
+ from transformers import NougatProcessor, VisionEncoderDecoderModel
8
+ import torch
9
+
10
+ processor = NougatProcessor.from_pretrained("nielsr/nougat")
11
+ model = VisionEncoderDecoderModel.from_pretrained("nielsr/nougat")
12
+
13
+ device = "cuda" if torch.cuda.is_available() else "cpu"
14
+ model.to(device)
15
+
16
+
17
+ def get_pdf(pdf_link):
18
+ unique_filename = f"{os.getcwd()}/downloaded_paper_{uuid.uuid4().hex}.pdf"
19
+
20
+ response = requests.get(pdf_link)
21
+
22
+ if response.status_code == 200:
23
+ with open(unique_filename, 'wb') as pdf_file:
24
+ pdf_file.write(response.content)
25
+ print("PDF downloaded successfully.")
26
+ else:
27
+ print("Failed to download the PDF.")
28
+ return unique_filename
29
+
30
+
31
+
32
+ def predict(image):
33
+ # prepare PDF image for the model
34
+ image = Image.open(image)
35
+ pixel_values = processor(image, return_tensors="pt").pixel_values
36
+
37
+ # generate transcription (here we only generate 30 tokens)
38
+ outputs = model.generate(
39
+ pixel_values.to(device),
40
+ min_length=1,
41
+ max_new_tokens=30,
42
+ bad_words_ids=[[processor.tokenizer.unk_token_id]],
43
+ )
44
+
45
+ sequence = processor.batch_decode(outputs, skip_special_tokens=True)[0]
46
+ sequence = processor.post_process_generation(sequence, fix_markdown=False)
47
+ return sequence
48
+
49
+
50
+
51
+ def inference(pdf_file, pdf_link):
52
+ if pdf_file is None:
53
+ if pdf_link == '':
54
+ print("No file is uploaded and No link is provided")
55
+ return "No data provided. Upload a pdf file or provide a pdf link and try again!"
56
+ else:
57
+ file_name = get_pdf(pdf_link)
58
+ else:
59
+ file_name = pdf_file.name
60
+ pdf_name = pdf_file.name.split('/')[-1].split('.')[0]
61
+
62
+ images = rasterize_paper(file_name, return_pil=True)
63
+ sequence = ""
64
+ # infer for every page and concat
65
+ for image in images:
66
+ sequence += predict(image)
67
+
68
+
69
+ content = sequence.replace(r'\(', '$').replace(r'\)', '$').replace(r'\[', '$$').replace(r'\]', '$$')
70
+ return content
71
+
72
+ import gradio as gr
73
+ import uuid
74
+ import os
75
+ import requests
76
+ import re
77
+
78
+ css = """
79
+ #mkd {
80
+ height: 500px;
81
+ overflow: auto;
82
+ border: 1px solid #ccc;
83
+ }
84
+ """
85
+
86
+ with gr.Blocks(css=css) as demo:
87
+ gr.HTML("<h1><center>Nougat: Neural Optical Understanding for Academic Documents 🍫<center><h1>")
88
+ gr.HTML("<h3><center>Lukas Blecher et al. <a href='https://arxiv.org/pdf/2308.13418.pdf' target='_blank'>Paper</a>, <a href='https://facebookresearch.github.io/nougat/'>Project</a><center></h3>")
89
+ gr.HTML("<h3><center>This demo is based on transformers implementation of Nougat 🤗<center><h3>")
90
+
91
+
92
+ with gr.Row():
93
+ mkd = gr.Markdown('<h4><center>Upload a PDF</center></h4>',scale=1)
94
+ mkd = gr.Markdown('<h4><center><i>OR</i></center></h4>',scale=1)
95
+ mkd = gr.Markdown('<h4><center>Provide a PDF link</center></h4>',scale=1)
96
+
97
+
98
+ with gr.Row():
99
+ mkd = gr.Markdown("Upload a PDF",scale=1)
100
+ mkd = gr.Markdown('OR',scale=1)
101
+ mkd = gr.Markdown('Provide a PDF link',scale=1)
102
+
103
+ with gr.Row(equal_height=True):
104
+ pdf_file = gr.File(label='PDF 📑', file_count='single', scale=1)
105
+ pdf_link = gr.Textbox(placeholder='Enter an arxiv link here', label='Link to Paper🔗', scale=1)
106
+
107
+ with gr.Row():
108
+ btn = gr.Button('Run Nougat 🍫')
109
+ clr = gr.Button('Clear 🧼')
110
+
111
+ output_headline = gr.Markdown("PDF converted to markup language through Nougat-OCR👇")
112
+ parsed_output = gr.Markdown(elem_id='mkd', value='OCR Output 📝')
113
+
114
+ btn.click(inference, [pdf_file, pdf_link], parsed_output )
115
+ clr.click(lambda : (gr.update(value=None),
116
+ gr.update(value=None),
117
+ gr.update(value=None)),
118
+ [],
119
+ [pdf_file, pdf_link, parsed_output]
120
+ )
121
+
122
+
123
+
124
+ demo.queue()
125
+ demo.launch(debug=True)