Spaces:
Sleeping
Sleeping
File size: 2,219 Bytes
d58981a e085c95 a0ead68 1872809 c379e84 a74ff63 c379e84 eebad8b c379e84 19d685b a74ff63 cf7f7d8 c379e84 cf7f7d8 27d29ad 06510d6 c379e84 27d29ad cf7f7d8 c379e84 eebad8b c379e84 eebad8b c379e84 eebad8b c379e84 eebad8b c379e84 eebad8b |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 |
import os
os.system('chmod 777 /tmp')
os.system('apt-get update -y')
os.system("apt-get install tesseract-ocr")
from fastapi import FastAPI, File, Request, UploadFile, Body, Depends, HTTPException
from fastapi.security.api_key import APIKeyHeader
from typing import Optional, Annotated
from fastapi.encoders import jsonable_encoder
from PIL import Image
from io import BytesIO
import pytesseract
from nltk.tokenize import sent_tokenize
from transformers import MarianMTModel, MarianTokenizer
API_KEY = os.environ.get("API_KEY")
app = FastAPI()
api_key_header = APIKeyHeader(name="api_key", auto_error=False)
def get_api_key(api_key: Optional[str] = Depends(api_key_header)):
if api_key is None or api_key != API_KEY:
raise HTTPException(status_code=401, detail="Unauthorized access")
return api_key
@app.post("/api/ocr", response_model=dict)
async def ocr(
api_key: str = Depends(get_api_key),
image: UploadFile = File(...),
# languages: list = Body(["eng"])
):
try:
content = await image.read()
image = Image.open(BytesIO(content))
# text = pytesseract.image_to_string(image, lang="+".join(languages))
text = pytesseract.image_to_string(image, lang = 'eng')
except Exception as e:
return {"error": str(e)}, 500
# return jsonable_encoder({"text": text})
return {"ImageText": text}
@app.post("/api/translate", response_model=dict)
async def translate(
api_key: str = Depends(get_api_key),
text: str = Body(...),
src: str = "en",
trg: str = "zh",
):
if api_key != API_KEY:
return {"error": "Invalid API key"}, 401
tokenizer, model = get_model(src, trg)
translated_text = ""
for sentence in sent_tokenize(text):
translated_sub = model.generate(**tokenizer(sentence, return_tensors="pt"))[0]
translated_text += tokenizer.decode(translated_sub, skip_special_tokens=True) + "\n"
return jsonable_encoder({"translated_text": translated_text})
def get_model(src: str, trg: str):
model_name = f"Helsinki-NLP/opus-mt-{src}-{trg}"
tokenizer = MarianTokenizer.from_pretrained(model_name)
model = MarianMTModel.from_pretrained(model_name)
return tokenizer, model
|