Zaravya / app.py
viswanani's picture
Update app.py
892b370 verified
# app.py
"""
Final Menu OCR -> Excel (Batch) Gradio app
- Robust handling of Gradio temp file paths
- Parses filename (<StoreName>_<StoreCode> <BranchName>.<ext>)
- Produces one Excel per image (mapped to template A..S, row 3 onward)
- Returns a ZIP of all Excels and also individual Excel files for download
"""
import gradio as gr
import pandas as pd
import pytesseract
from pytesseract import Output
import cv2
import re
import tempfile
import shutil
import os
import numpy as np
from PIL import Image
from io import BytesIO
from zipfile import ZipFile
from openpyxl import load_workbook
PRICE_REGEX = re.compile(r"(?:₹|Rs\.?|INR)?\s*([0-9]{1,6}(?:\.[0-9]{1,2})?)(?:\s*/-)?\s*$", flags=re.IGNORECASE)
CATEGORY_HINTS = ["maggi", "noodles", "pizza", "burger", "rice", "continental", "beverages", "coffee", "tea"]
DEFAULTS = {
"Active": "1",
"Priority": "",
"Image": "",
"Food type": "",
"NoOfMains": "1",
"OnlineName": "",
"AlternateClassification": "",
"ItemTaxInclusive": "0",
"TaxPct": "",
"BrandName": "",
"ClassificationCode": "",
"HSN Code": ""
}
def safe_read_bytes(uploaded_file):
"""
uploaded_file may be a Gradio temp-file object. We try reading from the .name path if present,
otherwise fallback to uploaded_file.read()
"""
if uploaded_file is None:
return None
# Try using the path if it exists (this handles /tmp/gradio/...)
try:
path = getattr(uploaded_file, "name", None)
if path and os.path.exists(path):
with open(path, "rb") as f:
return f.read()
except Exception:
pass
# fallback to reading the object itself
try:
uploaded_file.seek(0)
except Exception:
pass
try:
return uploaded_file.read()
except Exception:
return None
def get_original_basename(uploaded_file):
"""
Return basename from uploaded_file.name (works with Gradio temp paths)
"""
name_attr = getattr(uploaded_file, "name", "")
if not name_attr:
return "unknown"
return os.path.basename(name_attr)
def parse_filename(filename: str):
base = os.path.splitext(os.path.basename(filename))[0]
if "_" in base:
left, right = base.split("_", 1)
store_name = left.strip()
parts = right.strip().split(" ", 1)
store_code = parts[0].strip()
branch_name = parts[1].strip() if len(parts) > 1 else ""
else:
m = re.match(r"(.+?)\s*\((.+?)\)", base)
if m:
store_name = m.group(1).strip()
branch_name = m.group(2).strip()
store_code = ""
else:
store_name = base
store_code = ""
branch_name = ""
return store_name, store_code, branch_name
def preprocess_image(np_img):
gray = cv2.cvtColor(np_img, cv2.COLOR_RGB2GRAY)
h, w = gray.shape[:2]
if min(h, w) < 1000:
scale = max(1.5, 1000.0 / min(h, w))
gray = cv2.resize(gray, None, fx=scale, fy=scale, interpolation=cv2.INTER_CUBIC)
th = cv2.adaptiveThreshold(gray, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
cv2.THRESH_BINARY, 41, 11)
kernel = np.ones((1, 1), np.uint8)
opened = cv2.morphologyEx(th, cv2.MORPH_OPEN, kernel)
return opened
def ocr_with_confidence(pil_img):
try:
data = pytesseract.image_to_data(pil_img, output_type=Output.DICT, lang='eng')
except Exception as e:
raise RuntimeError(f"Tesseract OCR failed: {e}. Ensure Tesseract is installed on the host.")
texts = data.get('text', [])
confs = data.get('conf', [])
block_nums = data.get('block_num', [])
par_nums = data.get('par_num', [])
line_nums = data.get('line_num', [])
lines_map = {}
for t, c, b, p, l in zip(texts, confs, block_nums, par_nums, line_nums):
if t is None or str(t).strip() == "":
continue
key = f"{b}_{p}_{l}"
if key not in lines_map:
lines_map[key] = {"tokens": [], "confs": []}
lines_map[key]["tokens"].append(str(t))
try:
conf_val = float(c)
except:
conf_val = -1.0
if conf_val >= 0:
lines_map[key]["confs"].append(conf_val)
lines = []
for key in sorted(lines_map.keys(), key=lambda x: tuple(map(int, x.split("_")))):
tokens = lines_map[key]["tokens"]
confs_line = lines_map[key]["confs"]
text_line = " ".join(tokens).strip()
avg_conf = round(sum(confs_line)/len(confs_line),2) if confs_line else 0.0
lines.append({"line": text_line, "conf": avg_conf})
full_text = "\n".join([l["line"] for l in lines])
return full_text, lines
def split_lines(text: str):
cleaned = re.sub(r"[•·●\t]", " ", text)
cleaned = re.sub(r"[ ]{2,}", " ", cleaned)
return [ln.strip() for ln in cleaned.splitlines() if ln.strip()]
def looks_like_category(line: str):
low = line.lower()
if any(k in low for k in CATEGORY_HINTS):
return True
if not re.search(r"\d", line) and len(line.split()) <= 6:
return True
return False
def parse_menu_lines(lines):
rows = []
current_parent = ""
current_category = ""
for ln in lines:
if looks_like_category(ln):
if ln.isupper() or any(k in ln.lower() for k in CATEGORY_HINTS):
current_parent = ln.strip(":- ")
continue
else:
current_category = ln.strip(":- ")
continue
m = PRICE_REGEX.search(ln)
if m:
price = m.group(1).strip()
name_part = PRICE_REGEX.sub("", ln).strip(" -:.")
row = {
"Parent Category": current_parent,
"Category": current_category,
"Name": name_part,
"Item Code": "",
"Master Item Name": name_part,
"EAN Code": "",
"Price": price,
"Active": DEFAULTS["Active"],
"Priority": DEFAULTS["Priority"],
"Image": DEFAULTS["Image"],
"Food type": DEFAULTS["Food type"],
"NoOfMains": DEFAULTS["NoOfMains"],
"OnlineName": DEFAULTS["OnlineName"],
"AlternateClassification": DEFAULTS["AlternateClassification"],
"ItemTaxInclusive": DEFAULTS["ItemTaxInclusive"],
"TaxPct": DEFAULTS["TaxPct"],
"BrandName": DEFAULTS["BrandName"],
"ClassificationCode": DEFAULTS["ClassificationCode"],
"HSN Code": DEFAULTS["HSN Code"]
}
rows.append(row)
else:
if re.search(r"\d", ln):
name_part = ln.strip()
row = {
"Parent Category": current_parent,
"Category": current_category,
"Name": name_part,
"Item Code": "",
"Master Item Name": name_part,
"EAN Code": "",
"Price": "",
"Active": DEFAULTS["Active"],
"Priority": DEFAULTS["Priority"],
"Image": DEFAULTS["Image"],
"Food type": DEFAULTS["Food type"],
"NoOfMains": DEFAULTS["NoOfMains"],
"OnlineName": DEFAULTS["OnlineName"],
"AlternateClassification": DEFAULTS["AlternateClassification"],
"ItemTaxInclusive": DEFAULTS["ItemTaxInclusive"],
"TaxPct": DEFAULTS["TaxPct"],
"BrandName": DEFAULTS["BrandName"],
"ClassificationCode": DEFAULTS["ClassificationCode"],
"HSN Code": DEFAULTS["HSN Code"]
}
rows.append(row)
return rows
def fill_template_bytes(template_path, rows, store_name, store_code, branch_name):
wb = load_workbook(template_path)
ws = wb.active
ws["A1"] = store_name
ws["B1"] = store_code
ws["C1"] = branch_name
start_row = 3
r = start_row
for item in rows:
ws.cell(row=r, column=1, value=item.get("Parent Category",""))
ws.cell(row=r, column=2, value=item.get("Category",""))
ws.cell(row=r, column=3, value=item.get("Name",""))
ws.cell(row=r, column=4, value=item.get("Item Code",""))
ws.cell(row=r, column=5, value=item.get("Master Item Name",""))
ws.cell(row=r, column=6, value=item.get("EAN Code",""))
ws.cell(row=r, column=7, value=item.get("Price",""))
ws.cell(row=r, column=8, value=item.get("Active",""))
ws.cell(row=r, column=9, value=item.get("Priority",""))
ws.cell(row=r, column=10, value=item.get("Image",""))
ws.cell(row=r, column=11, value=item.get("Food type",""))
ws.cell(row=r, column=12, value=item.get("NoOfMains",""))
ws.cell(row=r, column=13, value=item.get("OnlineName",""))
ws.cell(row=r, column=14, value=item.get("AlternateClassification",""))
ws.cell(row=r, column=15, value=item.get("ItemTaxInclusive",""))
ws.cell(row=r, column=16, value=item.get("TaxPct",""))
ws.cell(row=r, column=17, value=item.get("BrandName",""))
ws.cell(row=r, column=18, value=item.get("ClassificationCode",""))
ws.cell(row=r, column=19, value=item.get("HSN Code",""))
r += 1
out = BytesIO()
wb.save(out)
out.seek(0)
return out
def sanitize_filename(name):
return re.sub(r"[^\w\-_\. ]", "_", name)
def process_batch(images, template):
if images is None or template is None:
raise gr.Error("Please upload images and a template file.")
tempdir = tempfile.mkdtemp()
generated_paths = []
for uploaded in images:
try:
orig_basename = get_original_basename(uploaded)
store_name, store_code, branch_name = parse_filename(orig_basename)
data = safe_read_bytes(uploaded)
if data is None:
raise RuntimeError("Could not read uploaded image bytes.")
pil = Image.open(BytesIO(data)).convert("RGB")
np_img = np.array(pil)
pre = preprocess_image(np_img)
pil_pre = Image.fromarray(pre)
full_text, lines_conf = ocr_with_confidence(pil_pre)
lines = split_lines(full_text)
rows = parse_menu_lines(lines)
out_buf = fill_template_bytes(template.name, rows, store_name, store_code, branch_name)
out_name = sanitize_filename(os.path.splitext(orig_basename)[0]) + ".xlsx"
out_path = os.path.join(tempdir, out_name)
with open(out_path, "wb") as f:
f.write(out_buf.read())
generated_paths.append(out_path)
except Exception as e:
err_name = sanitize_filename(os.path.splitext(get_original_basename(uploaded))[0]) + "_ERROR.txt"
err_path = os.path.join(tempdir, err_name)
with open(err_path, "w", encoding="utf-8") as ef:
ef.write(str(e))
generated_paths.append(err_path)
zip_path = os.path.join(tempdir, "Menu_Results.zip")
with ZipFile(zip_path, "w") as zf:
for p in generated_paths:
zf.write(p, arcname=os.path.basename(p))
return zip_path, generated_paths
with gr.Blocks() as demo:
gr.Markdown("## Menu OCR → Excel (Batch)\nUpload multiple images and an Excel template. The app will parse filename metadata, OCR the menu, and produce one Excel per image.")
with gr.Row():
images_in = gr.File(file_count="multiple", label="Upload menu images", file_types=["image"])
template_in = gr.File(file_count="single", label="Upload Excel template (.xlsx)", file_types=[".xlsx"])
run_btn = gr.Button("Process all images")
zip_out = gr.File(label="Download ZIP of all Excel outputs")
files_out = gr.File(label="Download individual Excel files (multiple)", file_count="multiple")
status = gr.Textbox(label="Status")
def run(images, template):
try:
zip_path, files = process_batch(images, template)
return zip_path, files, f"Processed {len(files)} files. Download ZIP or individual files."
except Exception as e:
return None, [], f"Error: {e}"
run_btn.click(fn=run, inputs=[images_in, template_in], outputs=[zip_out, files_out, status])
demo.launch()