File size: 8,674 Bytes
87b2711
db47c8b
7c99e50
1ca62bc
87b2711
f52050f
7c99e50
f52050f
db47c8b
ea69d06
6082f24
f52050f
 
 
7c99e50
6082f24
f52050f
2fecabe
6082f24
f52050f
 
 
 
 
 
 
 
ec4bd6c
 
f52050f
 
 
 
 
 
 
 
 
 
 
ec4bd6c
db47c8b
f52050f
 
 
807214f
 
f52050f
 
 
 
 
807214f
f52050f
807214f
f52050f
 
 
807214f
f52050f
 
 
 
 
 
 
 
 
 
7c99e50
f52050f
807214f
f52050f
 
 
 
 
 
 
807214f
 
f52050f
40a85e4
 
f52050f
7f5acb9
40a85e4
 
 
 
 
 
807214f
 
 
 
 
 
 
 
 
 
 
 
4a7bfda
 
807214f
db47c8b
 
807214f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7c99e50
6082f24
7c99e50
 
6082f24
25ad79c
1463298
 
25ad79c
807214f
6082f24
 
 
7c99e50
6082f24
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
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
import os
import tempfile
import gradio as gr
import numpy as np
from transformers import Qwen2VLForConditionalGeneration, AutoProcessor
from qwen_vl_utils import process_vision_info
import torch
from ast import literal_eval
from PIL import Image

# Load the model on the available device(s)
model = Qwen2VLForConditionalGeneration.from_pretrained(
    "Qwen/Qwen2-VL-7B-Instruct", torch_dtype="auto", device_map="auto"
)

# Load the processor
processor = AutoProcessor.from_pretrained("Qwen/Qwen2-VL-7B-Instruct")

# Define your prompts
other_benifits = '''Extract the following information in the given format:
        {'other_benefits_and_information': {
            '401k eru: {'This Period':'', 'Year-to-Date':''}},
            'quota summary':
                            {
                            'sick:': '',
                            'vacation:': '',
                            }
            'payment method': 'eg. Direct payment',
            'Amount': 'eg. 12.99'
                            }
'''

tax_deductions = '''Extract the following information in the given format:
                {
                'tax_deductions': {
                    'federal:': {
                        'withholding tax:': {'Amount':'', 'Year-To_Date':""},
                        'ee social security tax:': {'Amount':'', 'Year-To_Date':""},
                        'ee medicare tax:': {'Amount':'', 'Year-To_Date':""}},
                    'california:': {
                        'withholding tax:': {'Amount':'', 'Year-To_Date':""},
                        'ee disability tax:': {'Amount':'', 'Year-To-Date':""}}},
                }
'''

def process_function(image_path, prompt):
    messages = [
        {
            "role": "user",
            "content": [
                {
                    "type": "image",
                    "image": image_path,  # Use the file path here
                },
                {"type": "text", "text":  prompt},
            ],
        }
    ]
    # Preparation for inference
    text = processor.apply_chat_template(
        messages, tokenize=False, add_generation_prompt=True
    )
    image_inputs, video_inputs = process_vision_info(messages)
    inputs = processor(
        text=[text],
        images=image_inputs,
        videos=video_inputs,
        padding=True,
        return_tensors="pt",
    )
    inputs = inputs.to("cuda")
    # Inference: Generation of the output
    generated_ids = model.generate(**inputs, max_new_tokens=1500)
    generated_ids_trimmed = [
        out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
    ]
    output_text = processor.batch_decode(
        generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
    )

    # Handle output text to convert it into JSON
    try:
        almost_json = output_text[0].split('```\n')[-1].split('\n```')[0]

        json = literal_eval(almost_json)
    except:
        try:
            almost_json = output_text[0].split('```json\n')[-1].split('\n```')[0]
            json = literal_eval(almost_json)
        except:
            json = output_text[0]

    return json

def process_document(image):
    # Save the uploaded image to a temporary file
    with tempfile.NamedTemporaryFile(delete=False, suffix=".jpg") as tmp_file:
        image = Image.fromarray(image)  # Convert NumPy array to PIL Image
        image.save(tmp_file.name)  # Save the image to the temporary file
        image_path = tmp_file.name  # Get the path of the saved file

    # Process the image with your model
    one = process_function(image_path, other_benifits)
    two = process_function(image_path, tax_deductions)


    # Optionally, you can delete the temporary file after use
    os.remove(image_path)

    return one, two



# def process_document(image):
#     with tempfile.NamedTemporaryFile(delete=False, suffix=".jpg") as tmp_file:
#         image = Image.fromarray(image)  
#         image.save(tmp_file.name) 
#         image_path = tmp_file.name  


#         messages = [
#         {
#             "role": "user",
#             "content": [
#                 {
#                     "type": "image",
#                     "image": image_path, 
#                 },
#                 {"type": "text", "text":  '''Extract the following information in the given format:
#                 {
#                 'tax_deductions': {
#                     'federal:': {
#                         'withholding tax:': {'Amount':'', 'Year-To_Date':""},
#                         'ee social security tax:': {'Amount':'', 'Year-To_Date':""},
#                         'ee medicare tax:': {'Amount':'', 'Year-To_Date':""}},
#                     'california:': {
#                         'withholding tax:': {'Amount':'', 'Year-To_Date':""},
#                         'ee disability tax:': {'Amount':'', 'Year-To-Date':""}}},
#                 }'''},
#             ],
#         }
#     ]
#     text = processor.apply_chat_template(
#         messages, tokenize=False, add_generation_prompt=True
#     )
#     image_inputs, video_inputs = process_vision_info(messages)
#     inputs = processor(
#         text=[text],
#         images=image_inputs,
#         videos=video_inputs,
#         padding=True,
#         return_tensors="pt",
#     )
#     inputs = inputs.to("cuda")
#     generated_ids = model.generate(**inputs, max_new_tokens=1500)
#     generated_ids_trimmed = [
#         out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
#     ]
#     output_text = processor.batch_decode(
#         generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
#     )
#     try:
#         almost_json = output_text[0].split('```\n')[-1].split('\n```')[0]

#         json = literal_eval(almost_json)
#     except:
#         try:
#             almost_json = output_text[0].split('```json\n')[-1].split('\n```')[0]
#             json = literal_eval(almost_json)
#         except:
#             json = output_text[0]

#     messages = [
#         {
#             "role": "user",
#             "content": [
#                 {
#                     "type": "image",
#                     "image": image_path,  
#                 },
#                 {"type": "text", "text":  '''Extract the following information in the given format:
#         {'other_benefits_and_information': {
#             '401k eru: {'This Period':'', 'Year-to-Date':''}},
#             'quota summary':
#                             {
#                             'sick:': '',
#                             'vacation:': '',
#                             }
#             'payment method': 'eg. Direct payment',
#             'Amount': 'eg. 12.99'
#                             }'''},
#             ],
#         }
#     ]
#     text = processor.apply_chat_template(
#         messages, tokenize=False, add_generation_prompt=True
#     )
#     image_inputs, video_inputs = process_vision_info(messages)
#     inputs = processor(
#         text=[text],
#         images=image_inputs,
#         videos=video_inputs,
#         padding=True,
#         return_tensors="pt",
#     )
#     inputs = inputs.to("cuda")
#     # Inference: Generation of the output
#     generated_ids = model.generate(**inputs, max_new_tokens=1500)
#     generated_ids_trimmed = [
#         out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
#     ]
#     output_text = processor.batch_decode(
#         generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
#     )
#     try:
#         almost_json_2 = output_text[0].split('```\n')[-1].split('\n```')[0]

#         json_2 = literal_eval(almost_json_2)
#     except:
#         try:
#             almost_json_2 = output_text[0].split('```json\n')[-1].split('\n```')[0]
#             json_2 = literal_eval(almost_json_2)
#         except:
#             json_2 = output_text[0]

#     # json_op = {
#     #     "tax_deductions": json,
#     #     "other_benifits": json_2
#     # }
#     # # Optionally, you can delete the temporary file after use
#     os.remove(image_path)

#     return json, json_2

# Create Gradio interface
demo = gr.Interface(
    fn=process_document,
    inputs="image",  # Gradio will handle the image input
    outputs=[
        gr.JSON(label="Tax Deductions Information"),  # First output box with heading
        gr.JSON(label="Other Benefits and Information")  # Second output box with heading
    ],
    title="<div style='text-align: center;'>Information Extraction From PaySlip</div>",
    examples=[["Slip_1.jpg"], ["Slip_2.jpg"]],
    cache_examples=False
)

demo.launch()