Commit
·
f28d3f0
1
Parent(s):
1064ef9
update code
Browse files
app.py
CHANGED
@@ -1,39 +1,43 @@
|
|
1 |
import os
|
2 |
-
import
|
3 |
-
from fastapi import FastAPI, HTTPException, UploadFile, File
|
4 |
from fastapi.middleware.cors import CORSMiddleware
|
|
|
|
|
5 |
from pdf2image import convert_from_bytes
|
6 |
from PIL import Image
|
7 |
-
|
8 |
-
import
|
9 |
-
import gradio as gr
|
10 |
|
11 |
-
#
|
12 |
-
|
13 |
-
|
14 |
-
|
15 |
-
|
16 |
-
|
17 |
-
|
18 |
-
)
|
19 |
|
20 |
-
#
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
|
|
|
|
|
|
|
|
|
26 |
|
27 |
-
# -------------------------
|
28 |
-
# Regex Field Extractor
|
29 |
-
# -------------------------
|
30 |
def extract_fields_regex(text: str) -> dict:
|
31 |
-
|
32 |
-
text = re.sub(r"
|
33 |
-
text = re.sub(r"\n
|
|
|
|
|
34 |
|
35 |
patterns = {
|
36 |
"เลขที่ผู้เสียภาษี": r"(?:TAX\s*ID|เลขที่ผู้เสียภาษี)[\s:\-\.]*([\d]{10,13})",
|
|
|
|
|
37 |
"เลขที่ใบกำกับภาษี": r"(?:TAX\s*INV\.?|เลขที่ใบกำกับภาษี|ใบกำกับ)[\s:\-\.]*([\d]{8,20})",
|
38 |
"จำนวนเงิน": r"(?:AMOUNT\s*THB|จำนวนเงิน|รวมเงิน)[\s:\-\.]*([\d,]+\.\d{2})",
|
39 |
"ราคาต่อลิตร": r"(?:Baht\/Litr\.?|Bath\/Ltr\.?|ราคาต่อลิตร|ราคา\/ลิตร|ราคาน้ำมัน)[\s:\-\.]*([\d,]+\.\d{2})",
|
@@ -47,73 +51,71 @@ def extract_fields_regex(text: str) -> dict:
|
|
47 |
for field, pattern in patterns.items():
|
48 |
match = re.search(pattern, text, re.IGNORECASE)
|
49 |
results[field] = match.group(1).strip() if match else None
|
|
|
|
|
|
|
|
|
|
|
|
|
50 |
|
51 |
return results
|
52 |
|
53 |
-
# -------------------------
|
54 |
-
# PDF Handling
|
55 |
-
# -------------------------
|
56 |
def pdf_to_image(file_bytes: bytes) -> Image.Image:
|
57 |
images = convert_from_bytes(file_bytes)
|
58 |
-
return images[0] #
|
59 |
-
|
60 |
-
# -------------------------
|
61 |
-
# FastAPI App
|
62 |
-
# -------------------------
|
63 |
-
app = FastAPI()
|
64 |
-
|
65 |
-
# Optional: Allow all CORS (customize for security)
|
66 |
-
app.add_middleware(
|
67 |
-
CORSMiddleware,
|
68 |
-
allow_origins=["*"],
|
69 |
-
allow_methods=["*"],
|
70 |
-
allow_headers=["*"],
|
71 |
-
)
|
72 |
|
|
|
73 |
@app.post("/api/ocr_receipt")
|
74 |
-
async def ocr_receipt(
|
|
|
|
|
|
|
|
|
|
|
|
|
75 |
content = await file.read()
|
76 |
|
77 |
try:
|
78 |
-
#
|
79 |
if file.filename.lower().endswith(".pdf"):
|
80 |
image = pdf_to_image(content)
|
|
|
81 |
else:
|
82 |
-
|
83 |
|
84 |
-
text =
|
85 |
-
|
86 |
|
87 |
return {
|
88 |
"raw_ocr": text,
|
89 |
-
"extracted_fields":
|
90 |
}
|
91 |
|
92 |
except Exception as e:
|
93 |
raise HTTPException(status_code=500, detail=str(e))
|
94 |
|
95 |
-
#
|
96 |
-
|
97 |
-
# -------------------------
|
98 |
-
def gradio_interface(image_path):
|
99 |
if isinstance(image_path, str) and image_path.lower().endswith(".pdf"):
|
100 |
with open(image_path, "rb") as f:
|
101 |
image = pdf_to_image(f.read())
|
102 |
else:
|
103 |
image = image_path
|
104 |
|
105 |
-
|
|
|
106 |
extracted = extract_fields_regex(text)
|
107 |
return text, extracted
|
108 |
|
109 |
with gr.Blocks() as demo:
|
110 |
-
gr.Markdown("# 🧾
|
111 |
with gr.Row():
|
112 |
img = gr.Image(type="filepath", label="อัปโหลดไฟล์ PDF หรือรูปภาพ")
|
113 |
-
out_text = gr.Textbox(label="ข้อความทั้งหมด", lines=
|
114 |
-
out_fields = gr.JSON(label="
|
115 |
-
btn = gr.Button("
|
116 |
btn.click(fn=gradio_interface, inputs=img, outputs=[out_text, out_fields])
|
117 |
|
118 |
-
#
|
119 |
-
|
|
|
|
1 |
import os
|
2 |
+
from fastapi import FastAPI, HTTPException, Header, UploadFile, File
|
|
|
3 |
from fastapi.middleware.cors import CORSMiddleware
|
4 |
+
import gradio as gr
|
5 |
+
from typhoon_ocr import ocr_document
|
6 |
from pdf2image import convert_from_bytes
|
7 |
from PIL import Image
|
8 |
+
import re
|
9 |
+
from dotenv import load_dotenv
|
|
|
10 |
|
11 |
+
# --- Load environment variables from .env ---
|
12 |
+
load_dotenv()
|
13 |
+
|
14 |
+
# --- Config ---
|
15 |
+
API_KEY = os.getenv("API_KEY")
|
16 |
+
TYPHOON_API_KEY = os.getenv("TYPHOON_OCR_API_KEY")
|
17 |
+
TYPHOON_BASE_URL = os.getenv("TYPHOON_BASE_URL", "https://api.opentyphoon.ai/v1")
|
|
|
18 |
|
19 |
+
# --- FastAPI App ---
|
20 |
+
app = FastAPI()
|
21 |
+
|
22 |
+
# CORS (optional for public usage)
|
23 |
+
app.add_middleware(
|
24 |
+
CORSMiddleware,
|
25 |
+
allow_origins=["*"],
|
26 |
+
allow_methods=["*"],
|
27 |
+
allow_headers=["*"],
|
28 |
+
)
|
29 |
|
|
|
|
|
|
|
30 |
def extract_fields_regex(text: str) -> dict:
|
31 |
+
# Preprocess text
|
32 |
+
text = re.sub(r"<.*?>", "", text) # Strip tags
|
33 |
+
text = re.sub(r"\n+", "\n", text) # Collapse newlines
|
34 |
+
text = re.sub(r"\s{2,}", " ", text) # Collapse multiple spaces
|
35 |
+
text = re.sub(r"\t+", " ", text)
|
36 |
|
37 |
patterns = {
|
38 |
"เลขที่ผู้เสียภาษี": r"(?:TAX\s*ID|เลขที่ผู้เสียภาษี)[\s:\-\.]*([\d]{10,13})",
|
39 |
+
|
40 |
+
# Updated pattern for correct tax invoice number
|
41 |
"เลขที่ใบกำกับภาษี": r"(?:TAX\s*INV\.?|เลขที่ใบกำกับภาษี|ใบกำกับ)[\s:\-\.]*([\d]{8,20})",
|
42 |
"จำนวนเงิน": r"(?:AMOUNT\s*THB|จำนวนเงิน|รวมเงิน)[\s:\-\.]*([\d,]+\.\d{2})",
|
43 |
"ราคาต่อลิตร": r"(?:Baht\/Litr\.?|Bath\/Ltr\.?|ราคาต่อลิตร|ราคา\/ลิตร|ราคาน้ำมัน)[\s:\-\.]*([\d,]+\.\d{2})",
|
|
|
51 |
for field, pattern in patterns.items():
|
52 |
match = re.search(pattern, text, re.IGNORECASE)
|
53 |
results[field] = match.group(1).strip() if match else None
|
54 |
+
|
55 |
+
# Optional fallback if regex fails
|
56 |
+
# if not results["เลขที่ใบกำกับภาษี"]:
|
57 |
+
# match = re.search(r"TAX\s*INV\.?\s*</td>\s*<td>\s*([\d\-]+)", text, re.IGNORECASE)
|
58 |
+
# if match:
|
59 |
+
# results["เลขที่ใบกำกับภาษี"] = match.group(1).strip()
|
60 |
|
61 |
return results
|
62 |
|
|
|
|
|
|
|
63 |
def pdf_to_image(file_bytes: bytes) -> Image.Image:
|
64 |
images = convert_from_bytes(file_bytes)
|
65 |
+
return images[0] # First page only
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
66 |
|
67 |
+
# --- API Endpoint ---
|
68 |
@app.post("/api/ocr_receipt")
|
69 |
+
async def ocr_receipt(
|
70 |
+
file: UploadFile = File(...),
|
71 |
+
x_api_key: str | None = Header(None),
|
72 |
+
):
|
73 |
+
if API_KEY and x_api_key != API_KEY:
|
74 |
+
raise HTTPException(status_code=401, detail="Invalid API key")
|
75 |
+
|
76 |
content = await file.read()
|
77 |
|
78 |
try:
|
79 |
+
# Handle PDF and image
|
80 |
if file.filename.lower().endswith(".pdf"):
|
81 |
image = pdf_to_image(content)
|
82 |
+
raw_output = ocr_document(image, task_type="structure")
|
83 |
else:
|
84 |
+
raw_output = ocr_document(content, task_type="structure")
|
85 |
|
86 |
+
text = raw_output if isinstance(raw_output, str) else raw_output.get("text", "")
|
87 |
+
extracted = extract_fields_regex(text)
|
88 |
|
89 |
return {
|
90 |
"raw_ocr": text,
|
91 |
+
"extracted_fields": extracted,
|
92 |
}
|
93 |
|
94 |
except Exception as e:
|
95 |
raise HTTPException(status_code=500, detail=str(e))
|
96 |
|
97 |
+
# --- Gradio UI ---
|
98 |
+
def gradio_interface(image_path: str | Image.Image):
|
|
|
|
|
99 |
if isinstance(image_path, str) and image_path.lower().endswith(".pdf"):
|
100 |
with open(image_path, "rb") as f:
|
101 |
image = pdf_to_image(f.read())
|
102 |
else:
|
103 |
image = image_path
|
104 |
|
105 |
+
raw = ocr_document(image, task_type="structure")
|
106 |
+
text = raw if isinstance(raw, str) else raw.get("text", "")
|
107 |
extracted = extract_fields_regex(text)
|
108 |
return text, extracted
|
109 |
|
110 |
with gr.Blocks() as demo:
|
111 |
+
gr.Markdown("# 🧾 แปลงและตรวจสอบใบเสร็จ")
|
112 |
with gr.Row():
|
113 |
img = gr.Image(type="filepath", label="อัปโหลดไฟล์ PDF หรือรูปภาพ")
|
114 |
+
out_text = gr.Textbox(label="ข้อความทั้งหมด", lines=10)
|
115 |
+
out_fields = gr.JSON(label="ข้อความที่ดึงออกมา")
|
116 |
+
btn = gr.Button("ประมวลผลใบเสร็จ")
|
117 |
btn.click(fn=gradio_interface, inputs=img, outputs=[out_text, out_fields])
|
118 |
|
119 |
+
# --- Mount Gradio on FastAPI ---
|
120 |
+
# app = gr.mount_gradio_app(app, demo, path="/ui")
|
121 |
+
demo.launch(share=False)
|
app_bk.py
CHANGED
@@ -1,43 +1,39 @@
|
|
1 |
import os
|
2 |
-
|
|
|
3 |
from fastapi.middleware.cors import CORSMiddleware
|
4 |
-
import gradio as gr
|
5 |
-
from typhoon_ocr import ocr_document
|
6 |
from pdf2image import convert_from_bytes
|
7 |
from PIL import Image
|
8 |
-
import
|
9 |
-
|
10 |
-
|
11 |
-
# --- Load environment variables from .env ---
|
12 |
-
load_dotenv()
|
13 |
-
|
14 |
-
# --- Config ---
|
15 |
-
API_KEY = os.getenv("API_KEY")
|
16 |
-
TYPHOON_API_KEY = os.getenv("TYPHOON_OCR_API_KEY")
|
17 |
-
TYPHOON_BASE_URL = os.getenv("TYPHOON_BASE_URL", "https://api.opentyphoon.ai/v1")
|
18 |
-
|
19 |
-
# --- FastAPI App ---
|
20 |
-
app = FastAPI()
|
21 |
|
22 |
-
#
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
|
|
|
28 |
)
|
29 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
30 |
def extract_fields_regex(text: str) -> dict:
|
31 |
-
#
|
32 |
-
text = re.sub(r"
|
33 |
-
text = re.sub(r"\n
|
34 |
-
text = re.sub(r"\s{2,}", " ", text) # Collapse multiple spaces
|
35 |
-
text = re.sub(r"\t+", " ", text)
|
36 |
|
37 |
patterns = {
|
38 |
"เลขที่ผู้เสียภาษี": r"(?:TAX\s*ID|เลขที่ผู้เสียภาษี)[\s:\-\.]*([\d]{10,13})",
|
39 |
-
|
40 |
-
# Updated pattern for correct tax invoice number
|
41 |
"เลขที่ใบกำกับภาษี": r"(?:TAX\s*INV\.?|เลขที่ใบกำกับภาษี|ใบกำกับ)[\s:\-\.]*([\d]{8,20})",
|
42 |
"จำนวนเงิน": r"(?:AMOUNT\s*THB|จำนวนเงิน|รวมเงิน)[\s:\-\.]*([\d,]+\.\d{2})",
|
43 |
"ราคาต่อลิตร": r"(?:Baht\/Litr\.?|Bath\/Ltr\.?|ราคาต่อลิตร|ราคา\/ลิตร|ราคาน้ำมัน)[\s:\-\.]*([\d,]+\.\d{2})",
|
@@ -51,71 +47,73 @@ def extract_fields_regex(text: str) -> dict:
|
|
51 |
for field, pattern in patterns.items():
|
52 |
match = re.search(pattern, text, re.IGNORECASE)
|
53 |
results[field] = match.group(1).strip() if match else None
|
54 |
-
|
55 |
-
# Optional fallback if regex fails
|
56 |
-
# if not results["เลขที่ใบกำกับภาษี"]:
|
57 |
-
# match = re.search(r"TAX\s*INV\.?\s*</td>\s*<td>\s*([\d\-]+)", text, re.IGNORECASE)
|
58 |
-
# if match:
|
59 |
-
# results["เลขที่ใบกำกับภาษี"] = match.group(1).strip()
|
60 |
|
61 |
return results
|
62 |
|
|
|
|
|
|
|
63 |
def pdf_to_image(file_bytes: bytes) -> Image.Image:
|
64 |
images = convert_from_bytes(file_bytes)
|
65 |
-
return images[0] #
|
66 |
|
67 |
-
#
|
68 |
-
|
69 |
-
|
70 |
-
|
71 |
-
|
72 |
-
)
|
73 |
-
|
74 |
-
|
|
|
|
|
|
|
|
|
75 |
|
|
|
|
|
76 |
content = await file.read()
|
77 |
|
78 |
try:
|
79 |
-
#
|
80 |
if file.filename.lower().endswith(".pdf"):
|
81 |
image = pdf_to_image(content)
|
82 |
-
raw_output = ocr_document(image, task_type="structure")
|
83 |
else:
|
84 |
-
|
85 |
|
86 |
-
text =
|
87 |
-
|
88 |
|
89 |
return {
|
90 |
"raw_ocr": text,
|
91 |
-
"extracted_fields":
|
92 |
}
|
93 |
|
94 |
except Exception as e:
|
95 |
raise HTTPException(status_code=500, detail=str(e))
|
96 |
|
97 |
-
#
|
98 |
-
|
|
|
|
|
99 |
if isinstance(image_path, str) and image_path.lower().endswith(".pdf"):
|
100 |
with open(image_path, "rb") as f:
|
101 |
image = pdf_to_image(f.read())
|
102 |
else:
|
103 |
image = image_path
|
104 |
|
105 |
-
|
106 |
-
text = raw if isinstance(raw, str) else raw.get("text", "")
|
107 |
extracted = extract_fields_regex(text)
|
108 |
return text, extracted
|
109 |
|
110 |
with gr.Blocks() as demo:
|
111 |
-
gr.Markdown("# 🧾
|
112 |
with gr.Row():
|
113 |
img = gr.Image(type="filepath", label="อัปโหลดไฟล์ PDF หรือรูปภาพ")
|
114 |
-
out_text = gr.Textbox(label="ข้อความทั้งหมด", lines=
|
115 |
-
out_fields = gr.JSON(label="
|
116 |
-
btn = gr.Button("
|
117 |
btn.click(fn=gradio_interface, inputs=img, outputs=[out_text, out_fields])
|
118 |
|
119 |
-
#
|
120 |
-
|
121 |
-
demo.launch(share=False)
|
|
|
1 |
import os
|
2 |
+
import re
|
3 |
+
from fastapi import FastAPI, HTTPException, UploadFile, File
|
4 |
from fastapi.middleware.cors import CORSMiddleware
|
|
|
|
|
5 |
from pdf2image import convert_from_bytes
|
6 |
from PIL import Image
|
7 |
+
from transformers import pipeline
|
8 |
+
import torch
|
9 |
+
import gradio as gr
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
10 |
|
11 |
+
# -------------------------
|
12 |
+
# Load Hugging Face Model
|
13 |
+
# -------------------------
|
14 |
+
ocr_pipeline = pipeline(
|
15 |
+
task="document-question-answering",
|
16 |
+
model="scb10x/typhoon-ocr-7b",
|
17 |
+
device=0 if torch.cuda.is_available() else -1
|
18 |
)
|
19 |
|
20 |
+
# -------------------------
|
21 |
+
# OCR Wrapper
|
22 |
+
# -------------------------
|
23 |
+
def ocr_document(image):
|
24 |
+
result = ocr_pipeline(image)
|
25 |
+
return result[0]["generated_text"] if result and isinstance(result, list) else str(result)
|
26 |
+
|
27 |
+
# -------------------------
|
28 |
+
# Regex Field Extractor
|
29 |
+
# -------------------------
|
30 |
def extract_fields_regex(text: str) -> dict:
|
31 |
+
text = re.sub(r"<.*?>", "", text) # remove tags
|
32 |
+
text = re.sub(r"\s{2,}", " ", text) # collapse spaces
|
33 |
+
text = re.sub(r"\n{2,}", "\n", text) # collapse newlines
|
|
|
|
|
34 |
|
35 |
patterns = {
|
36 |
"เลขที่ผู้เสียภาษี": r"(?:TAX\s*ID|เลขที่ผู้เสียภาษี)[\s:\-\.]*([\d]{10,13})",
|
|
|
|
|
37 |
"เลขที่ใบกำกับภาษี": r"(?:TAX\s*INV\.?|เลขที่ใบกำกับภาษี|ใบกำกับ)[\s:\-\.]*([\d]{8,20})",
|
38 |
"จำนวนเงิน": r"(?:AMOUNT\s*THB|จำนวนเงิน|รวมเงิน)[\s:\-\.]*([\d,]+\.\d{2})",
|
39 |
"ราคาต่อลิตร": r"(?:Baht\/Litr\.?|Bath\/Ltr\.?|ราคาต่อลิตร|ราคา\/ลิตร|ราคาน้ำมัน)[\s:\-\.]*([\d,]+\.\d{2})",
|
|
|
47 |
for field, pattern in patterns.items():
|
48 |
match = re.search(pattern, text, re.IGNORECASE)
|
49 |
results[field] = match.group(1).strip() if match else None
|
|
|
|
|
|
|
|
|
|
|
|
|
50 |
|
51 |
return results
|
52 |
|
53 |
+
# -------------------------
|
54 |
+
# PDF Handling
|
55 |
+
# -------------------------
|
56 |
def pdf_to_image(file_bytes: bytes) -> Image.Image:
|
57 |
images = convert_from_bytes(file_bytes)
|
58 |
+
return images[0] # first page only
|
59 |
|
60 |
+
# -------------------------
|
61 |
+
# FastAPI App
|
62 |
+
# -------------------------
|
63 |
+
app = FastAPI()
|
64 |
+
|
65 |
+
# Optional: Allow all CORS (customize for security)
|
66 |
+
app.add_middleware(
|
67 |
+
CORSMiddleware,
|
68 |
+
allow_origins=["*"],
|
69 |
+
allow_methods=["*"],
|
70 |
+
allow_headers=["*"],
|
71 |
+
)
|
72 |
|
73 |
+
@app.post("/api/ocr_receipt")
|
74 |
+
async def ocr_receipt(file: UploadFile = File(...)):
|
75 |
content = await file.read()
|
76 |
|
77 |
try:
|
78 |
+
# Convert PDF to image or use image directly
|
79 |
if file.filename.lower().endswith(".pdf"):
|
80 |
image = pdf_to_image(content)
|
|
|
81 |
else:
|
82 |
+
image = Image.open(file.file)
|
83 |
|
84 |
+
text = ocr_document(image)
|
85 |
+
fields = extract_fields_regex(text)
|
86 |
|
87 |
return {
|
88 |
"raw_ocr": text,
|
89 |
+
"extracted_fields": fields,
|
90 |
}
|
91 |
|
92 |
except Exception as e:
|
93 |
raise HTTPException(status_code=500, detail=str(e))
|
94 |
|
95 |
+
# -------------------------
|
96 |
+
# Gradio UI
|
97 |
+
# -------------------------
|
98 |
+
def gradio_interface(image_path):
|
99 |
if isinstance(image_path, str) and image_path.lower().endswith(".pdf"):
|
100 |
with open(image_path, "rb") as f:
|
101 |
image = pdf_to_image(f.read())
|
102 |
else:
|
103 |
image = image_path
|
104 |
|
105 |
+
text = ocr_document(image)
|
|
|
106 |
extracted = extract_fields_regex(text)
|
107 |
return text, extracted
|
108 |
|
109 |
with gr.Blocks() as demo:
|
110 |
+
gr.Markdown("# 🧾 OCR ใบเสร็จ (Thai Receipt Scanner)")
|
111 |
with gr.Row():
|
112 |
img = gr.Image(type="filepath", label="อัปโหลดไฟล์ PDF หรือรูปภาพ")
|
113 |
+
out_text = gr.Textbox(label="ข้อความทั้งหมด", lines=12)
|
114 |
+
out_fields = gr.JSON(label="ฟิลด์ที่ดึงออกมา")
|
115 |
+
btn = gr.Button("ประมวลผล")
|
116 |
btn.click(fn=gradio_interface, inputs=img, outputs=[out_text, out_fields])
|
117 |
|
118 |
+
# For Hugging Face Spaces — no uvicorn needed
|
119 |
+
demo.launch()
|
|