File size: 4,492 Bytes
06d7744
4092f98
a36dcb9
86e26e9
 
4092f98
 
 
 
 
 
 
 
 
155e281
4092f98
86e26e9
 
 
4092f98
 
155e281
86e26e9
 
4092f98
 
 
 
86e26e9
4092f98
86e26e9
4092f98
 
 
 
 
 
 
 
 
86e26e9
 
 
 
 
 
 
 
 
4092f98
86e26e9
 
 
4092f98
 
86e26e9
4092f98
 
86e26e9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4092f98
86e26e9
 
 
 
 
4092f98
 
 
 
 
 
 
86e26e9
a4d7a24
86e26e9
 
 
 
 
 
 
 
4092f98
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
import os
os.system('pip3 install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cpu')

import gradio as gr
import numpy as np
from transformers import AutoModelForTokenClassification
from datasets.features import ClassLabel
from transformers import AutoProcessor
from datasets import Features, Sequence, ClassLabel, Value, Array2D, Array3D
import torch
from datasets import load_metric
from transformers import LayoutLMv3ForTokenClassification
from transformers.data.data_collator import default_data_collator


from transformers import AutoModelForTokenClassification
from datasets import load_dataset
from PIL import Image, ImageDraw, ImageFont


processor = AutoProcessor.from_pretrained("Ammar-alhaj-ali/LayoutLMv3-Fine-Tuning-FUNSD", apply_ocr=True)
model = AutoModelForTokenClassification.from_pretrained("Ammar-alhaj-ali/LayoutLMv3-Fine-Tuning-FUNSD")

# load image example
dataset = load_dataset("darentang/generated", split="test")
Image.open(dataset[2]["image_path"]).convert("RGB").save("img1.png")
Image.open(dataset[1]["image_path"]).convert("RGB").save("img2.png")
Image.open(dataset[0]["image_path"]).convert("RGB").save("img3.png")
# define id2label, label2color
labels = ['O', 'B-HEADER', 'I-HEADER', 'B-QUESTION', 'I-QUESTION', 'B-ANSWER', 'I-ANSWER']
id2label = {v: k for v, k in enumerate(labels)}
label2color = {
    "B-HEADER": 'red',
    "I-HEADER": 'green',
    "B-QUESTION": 'orange',
    "I-QUESTION": "blue",
    "B-ANSWER": 'gray',
    "I-ANSWERE": 'violet',
    "O": 'orange'
  } 

def unnormalize_box(bbox, width, height):
     return [
         width * (bbox[0] / 1000),
         height * (bbox[1] / 1000),
         width * (bbox[2] / 1000),
         height * (bbox[3] / 1000),
     ]


def iob_to_label(label):
    return label



def process_image(image):

    print(type(image))
    width, height = image.size

    # encode
    encoding = processor(image, truncation=True, return_offsets_mapping=True, return_tensors="pt")
    offset_mapping = encoding.pop('offset_mapping')

    # forward pass
    outputs = model(**encoding)

    # get predictions
    predictions = outputs.logits.argmax(-1).squeeze().tolist()
    token_boxes = encoding.bbox.squeeze().tolist()

    # only keep non-subword predictions
    is_subword = np.array(offset_mapping.squeeze().tolist())[:,0] != 0
    true_predictions = [id2label[pred] for idx, pred in enumerate(predictions) if not is_subword[idx]]
    true_boxes = [unnormalize_box(box, width, height) for idx, box in enumerate(token_boxes) if not is_subword[idx]]

    # draw predictions over the image
    draw = ImageDraw.Draw(image)
    font = ImageFont.load_default()
    for prediction, box in zip(true_predictions, true_boxes):
        predicted_label = iob_to_label(prediction)
        draw.rectangle(box, outline=label2color[predicted_label])
        draw.text((box[0]+10, box[1]-10), text=predicted_label, fill=label2color[predicted_label], font=font)
    
    return image


title = "Extraction d'informations de factures en utilisant le modèle LayoutLMv3"
description = "J'utilise LayoutLMv3 de Microsoft formé sur un ensemble de données de factures pour prédire le nom de l'émetteur de factures, l'adresse de l'émetteur de factures, le code postal de l'émetteur de factures, la date d'échéance, la TPS, la date de facturation, le numéro de facture, le sous-total et le total. Pour l'utiliser, il suffit de télécharger une image ou d'utiliser l'exemple d'image ci-dessous. Les résultats seront affichés en quelques secondes."

article="<b>References</b><br>[1] Y. Xu et al., “LayoutLMv3: Pre-training for Document AI with Unified Text and Image Masking.” 2022. <a href='https://arxiv.org/abs/2204.08387'>Paper Link</a><br>[2]  <a href='https://github.com/NielsRogge/Transformers-Tutorials/tree/master/LayoutLMv3'>LayoutLMv3 training and inference</a>" 

examples =[['img1.png'],['img2.png'],['img3.png']]

css = """.output_image, .input_image {height: 600px !important}"""

iface = gr.Interface(fn=process_image, 
                     inputs=gr.inputs.Image(type="pil"), 
                     outputs=gr.outputs.Image(type="pil", label="annotated image"),
                     title=title,
                     description=description,
                     article=article,
                     examples=examples,
                     css=css,
                     analytics_enabled = True, enable_queue=True)

iface.launch(inline=False, share=False, debug=False)