imageocr / main.py
sarangg's picture
Switch to Google Vision only - remove CLIP/transformers/torch
3004157
Raw History Blame Contribute Delete
6.8 kB
from fastapi import FastAPI, Depends, HTTPException, status, UploadFile, File
from fastapi.middleware.cors import CORSMiddleware
from fastapi.responses import FileResponse
from fastapi.security import APIKeyHeader
import io
import os
import base64
import requests
from dotenv import load_dotenv
# Load .env file for local development
load_dotenv()
app = FastAPI(title="Image OCR & Vision API")
app.add_middleware(
CORSMiddleware,
allow_origins=["*"],
allow_credentials=True,
allow_methods=["*"],
allow_headers=["*"],
)
# ── Lazy OCR model ─────────────────────────────────────────────────────────────
_ocr_reader = None
def get_ocr_reader():
global _ocr_reader
if _ocr_reader is None:
import easyocr
print("Loading EasyOCR model...")
_ocr_reader = easyocr.Reader(['en'])
print("EasyOCR ready!")
return _ocr_reader
# ── Config ─────────────────────────────────────────────────────────────────────
API_KEY = os.environ.get("API_KEY", "xkLm9Qz4BvWtUa2Nf8PjX1")
API_KEY_NAME = "x-api-key"
api_key_header = APIKeyHeader(name=API_KEY_NAME, auto_error=False)
GOOGLE_VISION_API_KEY = os.environ.get("GOOGLE_VISION_API_KEY", "")
def get_api_key(api_key: str = Depends(api_key_header)):
if api_key == API_KEY:
return api_key
raise HTTPException(status_code=status.HTTP_401_UNAUTHORIZED, detail="Invalid or missing API Key")
# ── OCR helpers ─────────────────────────────────────────────────────────────────
def extract_text_from_image_bytes(image_bytes: bytes) -> str:
reader = get_ocr_reader()
result = reader.readtext(image_bytes, detail=0)
return "\n".join(result)
def process_pdf(pdf_bytes: bytes) -> str:
import fitz
doc = fitz.open(stream=pdf_bytes, filetype="pdf")
pages = []
for i in range(len(doc)):
page = doc.load_page(i)
pix = page.get_pixmap(matrix=fitz.Matrix(2, 2))
text = extract_text_from_image_bytes(pix.tobytes("png"))
if text.strip():
pages.append(f"--- Page {i+1} ---\n{text}")
doc.close()
return "\n\n".join(pages)
# ── Google Vision API ───────────────────────────────────────────────────────────
def classify_with_google(image_bytes: bytes) -> dict:
"""
Calls Google Vision API to detect labels, objects and web entities.
Returns a structured result with category and labels.
"""
if not GOOGLE_VISION_API_KEY:
raise HTTPException(
status_code=503,
detail="Google Vision API key is not configured. Add GOOGLE_VISION_API_KEY to your environment secrets."
)
encoded = base64.b64encode(image_bytes).decode("utf-8")
payload = {
"requests": [{
"image": {"content": encoded},
"features": [
{"type": "LABEL_DETECTION", "maxResults": 10},
{"type": "OBJECT_LOCALIZATION","maxResults": 5},
{"type": "WEB_DETECTION", "maxResults": 5},
]
}]
}
url = f"https://vision.googleapis.com/v1/images:annotate?key={GOOGLE_VISION_API_KEY}"
resp = requests.post(url, json=payload, timeout=20)
if resp.status_code == 400:
raise HTTPException(status_code=400, detail="Invalid Google Vision API key or request.")
if resp.status_code in (403, 429):
raise HTTPException(status_code=429, detail="Google Vision quota exceeded or API not enabled.")
resp.raise_for_status()
data = resp.json()["responses"][0]
labels = [l["description"] for l in data.get("labelAnnotations", [])]
objects = [o["name"] for o in data.get("localizedObjectAnnotations", [])]
# Web detection gives best product names
web_data = data.get("webDetection", {})
web_entities = [e["description"] for e in web_data.get("webEntities", []) if e.get("description")]
best_guess = [g["label"] for g in web_data.get("bestGuessLabels", [])]
# Combine: best guess > web entities > objects > labels
all_labels = list(dict.fromkeys(best_guess + web_entities + objects + labels))
category = all_labels[0] if all_labels else "Unknown"
return {
"labels": all_labels[:10], # top 10
"category": category,
"source": "google_vision"
}
# ── Routes ──────────────────────────────────────────────────────────────────────
@app.get("/", tags=["UI"])
async def serve_ui():
return FileResponse("index.html")
@app.post("/ocr", tags=["OCR"])
async def perform_ocr(
api_key: str = Depends(get_api_key),
file: UploadFile = File(...)
):
"""Extract text from an image or PDF."""
contents = await file.read()
filename = file.filename or ""
content_type = file.content_type or ""
is_pdf = content_type == "application/pdf" or filename.lower().endswith(".pdf")
is_image = content_type.startswith("image/")
if not is_pdf and not is_image:
raise HTTPException(status_code=400, detail="Upload an image or PDF.")
try:
text = process_pdf(contents) if is_pdf else extract_text_from_image_bytes(contents)
file_type = "pdf" if is_pdf else "image"
return {"success": True, "filename": filename, "file_type": file_type, "extracted_text": text.strip()}
except Exception as e:
raise HTTPException(status_code=500, detail=f"OCR failed: {str(e)}")
@app.post("/classify", tags=["Image Classification"])
async def classify_image(
api_key: str = Depends(get_api_key),
file: UploadFile = File(...)
):
"""
Identify a product or object in an image using Google Vision AI.
Requires GOOGLE_VISION_API_KEY to be set in environment secrets.
"""
if not file.content_type.startswith("image/"):
raise HTTPException(status_code=400, detail="Please upload an image file.")
contents = await file.read()
result = classify_with_google(contents)
return {"success": True, "filename": file.filename, **result}
if __name__ == "__main__":
import uvicorn
port = int(os.environ.get("PORT", 4002))
uvicorn.run("main:app", host="0.0.0.0", port=port, reload=True)