Files
odoo-ai/agent_service/tools/receipt_parser.py
Carlos Garcia 69519393c1 Add EasyOCR engine for receipt image parsing
EasyOCR (deep-learning OCR) replaces Tesseract as the default engine for
receipt images. It handles phone photos, thermal paper, dot-matrix fonts,
and rotated images significantly better than Tesseract without requiring
manual preprocessing pipelines.

Key design decisions:
- OCR_ENGINE=easyocr (default) | tesseract — switchable via .env, no rebuild
- EasyOCR Reader is a module-level singleton: model loaded once per container
  start, not per receipt
- Falls back to Tesseract automatically if EasyOCR fails or returns < 20 chars
- EXIF rotation fix still applied before EasyOCR (phone photo orientation)
- Images resized to max 2000px width for speed before passing to EasyOCR
- _easyocr_to_text() groups detections into visual lines (y-overlap) and
  sorts left-to-right within each line for clean single-string output

Revert: echo "OCR_ENGINE=tesseract" >> .env && docker compose up -d agent-service

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-05-21 01:22:22 -04:00

360 lines
14 KiB
Python

from __future__ import annotations
import base64
import hashlib
import io
import logging
import os
import re
import zipfile
from pathlib import Path
logger = logging.getLogger(__name__)
# Extract YYYYMMDD from filenames like 20260509_180857.jpg
_DATE_PATTERN = re.compile(r'(\d{4})(\d{2})(\d{2})_\d{6}')
# ---------------------------------------------------------------------------
# OCR engine selection
# ---------------------------------------------------------------------------
# Set OCR_ENGINE=tesseract in .env to revert to the old Tesseract pipeline.
# Default is easyocr which handles phone photos and difficult fonts better.
def _get_ocr_engine() -> str:
return os.environ.get('OCR_ENGINE', 'easyocr').lower()
# EasyOCR Reader is expensive to initialise (~10-30s on first call while it
# loads model weights). Cache it as a module-level singleton so the cost is
# paid once per container start, not once per receipt.
_easyocr_reader = None
def _get_easyocr_reader():
global _easyocr_reader
if _easyocr_reader is None:
import easyocr
logger.info('EasyOCR: initialising reader (first use — loading model weights)')
_easyocr_reader = easyocr.Reader(['en'], verbose=False)
logger.info('EasyOCR: reader ready')
return _easyocr_reader
_MIME = {
'.jpg': 'image/jpeg', '.jpeg': 'image/jpeg',
'.png': 'image/png', '.gif': 'image/gif',
'.bmp': 'image/bmp', '.tiff': 'image/tiff', '.tif': 'image/tiff',
'.webp': 'image/webp', '.pdf': 'application/pdf',
'.html': 'text/html', '.htm': 'text/html',
'.txt': 'text/plain', '.zip': 'application/zip',
}
_IMAGE_EXTS = {'.jpg', '.jpeg', '.png', '.gif', '.bmp', '.tiff', '.tif', '.webp'}
def parse_upload(filename: str, data: bytes) -> list[dict]:
"""
Parse one uploaded file into a list of receipt dicts.
ZIP files are recursively unpacked; all other types return a single entry.
Each dict: {filename, text, b64, mimetype}
"""
ext = Path(filename).suffix.lower()
if ext == '.zip':
return _extract_zip(filename, data)
b64 = base64.b64encode(data).decode()
mimetype = _MIME.get(ext, 'application/octet-stream')
sha256 = hashlib.sha256(data).hexdigest()
# Extract date from timestamp-style filenames (e.g. 20260509_180857.jpg)
date_from_name = None
m = _DATE_PATTERN.search(filename)
if m:
date_from_name = f'{m.group(1)}-{m.group(2)}-{m.group(3)}'
if ext in _IMAGE_EXTS:
text = _ocr_image(data, filename)
elif ext == '.pdf':
text = _extract_pdf(data, filename)
elif ext in ('.html', '.htm'):
text = _extract_html(data, filename)
elif ext == '.txt':
text = data.decode('utf-8', errors='replace')
else:
try:
text = data.decode('utf-8', errors='replace')
except Exception:
text = f'[Binary file: {filename}]'
return [{'filename': filename, 'text': text, 'b64': b64, 'mimetype': mimetype,
'sha256': sha256, 'date_from_name': date_from_name}]
def _extract_zip(zip_filename: str, data: bytes) -> list[dict]:
results = []
try:
with zipfile.ZipFile(io.BytesIO(data)) as zf:
for member in zf.namelist():
if member.endswith('/'):
continue
try:
member_data = zf.read(member)
results.extend(parse_upload(Path(member).name, member_data))
except Exception as exc:
logger.warning('receipt_parser: zip member %s failed: %s', member, exc)
except Exception as exc:
logger.error('receipt_parser: zip %s failed: %s', zip_filename, exc)
return results
def _ocr_image(data: bytes, filename: str) -> str:
"""Dispatch to the configured OCR engine (EasyOCR or Tesseract)."""
if _get_ocr_engine() == 'easyocr':
return _ocr_image_easyocr(data, filename)
return _ocr_image_tesseract(data, filename)
def _easyocr_to_text(results: list) -> str:
"""Convert EasyOCR result list to a single text string.
EasyOCR returns a list of (bbox, text, confidence) tuples. We filter
low-confidence detections, sort top-to-bottom then left-to-right, and
join with newlines. Receipt images are typically single-column so a
simple y-sort produces a clean reading order.
Adjacent words on the same horizontal band (y within 40% of the tallest
box's height in that group) are merged onto one line — this keeps a
label like "TOTAL 42.90" on a single line instead of two lines,
which is important for the labeled-total regex in expenses_agent.py.
"""
if not results:
return ''
# Filter and extract geometry
boxes = []
for bbox, text, conf in results:
if conf < 0.3 or not text.strip():
continue
ys = [pt[1] for pt in bbox]
xs = [pt[0] for pt in bbox]
boxes.append({
'y_top': min(ys), 'y_bot': max(ys),
'x_left': min(xs), 'text': text.strip(),
})
if not boxes:
return ''
boxes.sort(key=lambda b: (b['y_top'], b['x_left']))
# Group into visual lines
lines: list[list[dict]] = []
current: list[dict] = [boxes[0]]
for box in boxes[1:]:
# Compute the current line's y-span
cy_top = min(b['y_top'] for b in current)
cy_bot = max(b['y_bot'] for b in current)
height = max(cy_bot - cy_top, 1)
# This box belongs to the same line if its top overlaps the current band
if box['y_top'] < cy_bot - height * 0.3:
current.append(box)
else:
lines.append(sorted(current, key=lambda b: b['x_left']))
current = [box]
lines.append(sorted(current, key=lambda b: b['x_left']))
return '\n'.join(' '.join(b['text'] for b in line) for line in lines)
def _ocr_image_easyocr(data: bytes, filename: str) -> str:
"""EasyOCR pipeline — better than Tesseract on phone photos, thermal paper,
dot-matrix, and rotated receipts. Falls back to Tesseract on any error.
"""
try:
import numpy as np
from PIL import Image, ImageOps
reader = _get_easyocr_reader()
img = Image.open(io.BytesIO(data))
# EXIF rotation — same fix as the Tesseract pipeline
try:
img = ImageOps.exif_transpose(img)
except Exception:
pass
# Resize very large images for speed; EasyOCR is accurate but slow on
# images wider than ~2000px (typical 12MP phone photo is ~4000px wide).
max_w = 2000
if img.width > max_w:
scale = max_w / img.width
img = img.resize((max_w, int(img.height * scale)), Image.LANCZOS)
# EasyOCR accepts a numpy array directly
img_array = np.array(img)
results = reader.readtext(img_array)
text = _easyocr_to_text(results)
logger.debug('EasyOCR %s: %d chars', filename, len(text))
if len(text) >= 20:
return text
# Very short result — try Tesseract as fallback before giving up
logger.warning('EasyOCR %s: only %d chars, trying Tesseract fallback',
filename, len(text))
tess = _ocr_image_tesseract(data, filename)
return tess if len(tess) > len(text) else text
except ImportError:
logger.warning('easyocr/numpy not installed — falling back to Tesseract for %s', filename)
return _ocr_image_tesseract(data, filename)
except Exception as exc:
logger.warning('EasyOCR failed for %s: %s — falling back to Tesseract', filename, exc)
return _ocr_image_tesseract(data, filename)
def _ocr_image_tesseract(data: bytes, filename: str) -> str:
"""Tesseract-based OCR pipeline with phone-photo preprocessing."""
try:
from PIL import Image, ImageFilter, ImageOps
import pytesseract
img = Image.open(io.BytesIO(data))
# ── Step 1: EXIF rotation correction ─────────────────────────────────
# Phone photos are stored with EXIF orientation metadata but the pixel
# data is not actually rotated. Without this fix Tesseract reads a
# portrait receipt as a landscape image and produces garbage.
try:
img = ImageOps.exif_transpose(img)
except Exception:
pass # exif_transpose requires Pillow >= 6.0
# ── Step 1b: Content-based rotation correction ───────────────────────
# EXIF transpose (Step 1) only corrects for phone-tilt metadata.
# If the receipt was physically laid sideways in the frame (e.g. a
# landscape receipt photographed with the phone upright), the pixels
# are genuinely rotated and EXIF can't help. Ask Tesseract's OSD
# engine to detect the text orientation and rotate to correct it.
try:
osd = pytesseract.image_to_osd(img, config='--psm 0')
_am = re.search(r'Rotate:\s*(\d+)', osd)
if _am:
_angle = int(_am.group(1))
if _angle:
img = img.rotate(_angle, expand=True)
logger.debug('OSD: rotated %s by %d°', filename, _angle)
except Exception:
pass # OSD unavailable or not enough text — proceed without correction
# ── Step 2: Resize to working width (1800px) ──────────────────────────
max_w = 1800
if img.width > max_w:
scale = max_w / img.width
img = img.resize((max_w, int(img.height * scale)), Image.LANCZOS)
# Upscale very small images — Tesseract accuracy drops below ~600px
elif img.width < 600:
scale = 600 / img.width
img = img.resize((600, int(img.height * scale)), Image.LANCZOS)
# ── Step 3: Grayscale + contrast ─────────────────────────────────────
img = ImageOps.grayscale(img)
img = ImageOps.autocontrast(img)
img_gray = img # save grayscale for fallback — before binarization
# ── Step 4: Sharpen then binarize ─────────────────────────────────────
# Sharpen first so edges are crisp before thresholding.
# Threshold 160 (was 140) — gentler for faint thermal-print receipts
# where light gray text would be wiped out by the stricter threshold.
img = img.filter(ImageFilter.SHARPEN)
img = img.point(lambda x: 0 if x < 160 else 255)
# ── Step 5: OCR — try PSM modes best-suited for receipt layout ────────
# PSM 6 = single uniform text block (best for single-column receipts)
# PSM 4 = single column, variable text sizes (wider fallback)
# PSM 11 = sparse text — last resort for badly segmented images
for psm in (6, 4, 11):
try:
text = pytesseract.image_to_string(
img, config=f'--oem 3 --psm {psm}').strip()
if len(text) >= 20:
logger.debug('Tesseract OCR %s: psm=%d %d chars', filename, psm, len(text))
return text
except Exception:
pass
# ── Step 5b: Grayscale fallback ───────────────────────────────────────
# Binarization at threshold 160 can destroy dot-matrix and certain
# thermal-print fonts (e.g. parking kiosk receipts) where character
# pixels are close to the threshold and get wiped to white. If every
# binarized attempt failed, retry on the plain grayscale image —
# Tesseract handles grey-level input reasonably well for these cases.
for psm in (6, 4, 11):
try:
text = pytesseract.image_to_string(
img_gray, config=f'--oem 3 --psm {psm}').strip()
if len(text) >= 20:
logger.debug('Tesseract grayscale fallback %s: psm=%d %d chars',
filename, psm, len(text))
return text
except Exception:
pass
logger.warning('Tesseract OCR %s: all PSM modes returned < 20 chars', filename)
return ''
except ImportError:
logger.warning('pytesseract/Pillow not installed — OCR unavailable for %s', filename)
return f'[Image: {filename} — install pytesseract+Pillow for OCR]'
except Exception as exc:
logger.warning('Tesseract OCR failed for %s: %s', filename, exc)
return f'[Image: {filename} — OCR failed: {exc}]'
def _extract_pdf(data: bytes, filename: str) -> str:
try:
import pdfplumber
parts = []
with pdfplumber.open(io.BytesIO(data)) as pdf:
for page in pdf.pages:
t = page.extract_text()
if t:
parts.append(t)
return '\n'.join(parts).strip()
except ImportError:
logger.warning('pdfplumber not installed — PDF extraction unavailable for %s', filename)
return f'[PDF: {filename} — install pdfplumber for text extraction]'
except Exception as exc:
logger.warning('PDF extraction failed for %s: %s', filename, exc)
return f'[PDF: {filename} — extraction failed: {exc}]'
def _extract_html(data: bytes, filename: str) -> str:
try:
from html.parser import HTMLParser
class _TextExtractor(HTMLParser):
def __init__(self):
super().__init__()
self._parts: list[str] = []
self._skip = False
def handle_starttag(self, tag, attrs):
if tag in ('script', 'style'):
self._skip = True
def handle_endtag(self, tag):
if tag in ('script', 'style'):
self._skip = False
def handle_data(self, data):
if not self._skip:
s = data.strip()
if s:
self._parts.append(s)
def text(self):
return ' '.join(self._parts)
parser = _TextExtractor()
parser.feed(data.decode('utf-8', errors='replace'))
return parser.text()
except Exception as exc:
logger.warning('HTML extraction failed for %s: %s', filename, exc)
return f'[HTML: {filename} — extraction failed: {exc}]'