Update app.py
Browse files
app.py
CHANGED
@@ -1,23 +1,23 @@
|
|
1 |
import gradio as gr
|
2 |
from paddleocr import PaddleOCR
|
3 |
import numpy as np
|
4 |
-
import openai
|
5 |
import os
|
6 |
from langdetect import detect
|
|
|
|
|
7 |
|
8 |
# Initialize PaddleOCR
|
9 |
ocr_reader = PaddleOCR(use_angle_cls=True, lang='en')
|
10 |
|
11 |
# Initialize Whisper Model via Hugging Face Transformers
|
12 |
-
from transformers import pipeline
|
13 |
whisper_model = pipeline(
|
14 |
task="automatic-speech-recognition",
|
15 |
model="openai/whisper-small",
|
16 |
device=0
|
17 |
)
|
18 |
|
19 |
-
#
|
20 |
-
|
21 |
|
22 |
def detect_language(text):
|
23 |
try:
|
@@ -31,7 +31,7 @@ def gpt_clean_and_translate(text, target_language):
|
|
31 |
return "No text detected.", ""
|
32 |
|
33 |
prompt = f"""
|
34 |
-
You are an expert document reader and translator. You will receive
|
35 |
|
36 |
1. Identify and extract these fields: Name, Address, Date of Birth, Expiry Date, Class, Sex.
|
37 |
2. Output the information in full English sentences.
|
@@ -39,7 +39,7 @@ You are an expert document reader and translator. You will receive a noisy extra
|
|
39 |
If the target language is English, just output clean English sentences.
|
40 |
"""
|
41 |
|
42 |
-
response =
|
43 |
model="gpt-4o",
|
44 |
messages=[
|
45 |
{"role": "system", "content": prompt},
|
@@ -48,7 +48,7 @@ If the target language is English, just output clean English sentences.
|
|
48 |
temperature=0.2
|
49 |
)
|
50 |
|
51 |
-
cleaned_translation = response
|
52 |
return cleaned_translation
|
53 |
|
54 |
def process_document(image, target_language, language_group):
|
|
|
1 |
import gradio as gr
|
2 |
from paddleocr import PaddleOCR
|
3 |
import numpy as np
|
|
|
4 |
import os
|
5 |
from langdetect import detect
|
6 |
+
from openai import OpenAI
|
7 |
+
from transformers import pipeline
|
8 |
|
9 |
# Initialize PaddleOCR
|
10 |
ocr_reader = PaddleOCR(use_angle_cls=True, lang='en')
|
11 |
|
12 |
# Initialize Whisper Model via Hugging Face Transformers
|
|
|
13 |
whisper_model = pipeline(
|
14 |
task="automatic-speech-recognition",
|
15 |
model="openai/whisper-small",
|
16 |
device=0
|
17 |
)
|
18 |
|
19 |
+
# Initialize OpenAI Client
|
20 |
+
client = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))
|
21 |
|
22 |
def detect_language(text):
|
23 |
try:
|
|
|
31 |
return "No text detected.", ""
|
32 |
|
33 |
prompt = f"""
|
34 |
+
You are an expert document reader and translator. You will receive noisy extracted text from a government ID. Your tasks:
|
35 |
|
36 |
1. Identify and extract these fields: Name, Address, Date of Birth, Expiry Date, Class, Sex.
|
37 |
2. Output the information in full English sentences.
|
|
|
39 |
If the target language is English, just output clean English sentences.
|
40 |
"""
|
41 |
|
42 |
+
response = client.chat.completions.create(
|
43 |
model="gpt-4o",
|
44 |
messages=[
|
45 |
{"role": "system", "content": prompt},
|
|
|
48 |
temperature=0.2
|
49 |
)
|
50 |
|
51 |
+
cleaned_translation = response.choices[0].message.content.strip()
|
52 |
return cleaned_translation
|
53 |
|
54 |
def process_document(image, target_language, language_group):
|