app.py

import gradio as gr
import requests
import gtts as gt
from PIL import Image
from gradio_client import Client
from googletrans import Translator
import cv2
import numpy as np
import tempfile
import base64
from io import BytesIO 

def trans(text, lang='ta'):
    translator = Translator()
    out = translator.translate(text, dest=lang)
    tts = gt.gTTS(text=out.text, lang=lang)
    # Save the audio as a temporary file
    temp_audio_file = tempfile.NamedTemporaryFile(suffix=".mp3", delete=False)
    tts.save(temp_audio_file.name)
    return temp_audio_file.name

def object_recognition(image_array, lang):
    # Convert the NumPy array to PIL Image
    image = Image.fromarray(image_array)

    API_URL = "https://api-inference.huggingface.co/models/Salesforce/blip-image-captioning-large"
    headers = {"Authorization": "Bearer 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"}

    with open("temp_image.jpg", "wb") as f:
        image.save(f, format="JPEG")

    with open("temp_image.jpg", "rb") as f:
        response = requests.post(API_URL, headers=headers, data=f)

    output = response.json()
    result = output[0]['generated_text']
    text = "Object recognition result for the captured image."
    audio_file = trans(result, lang)
    return audio_file

def ocr_detection(image_array, lang):
    image = Image.fromarray(image_array)

    buffered = BytesIO()
    image.save(buffered, format="PNG")
    image_base64 = base64.b64encode(buffered.getvalue()).decode()

    response = requests.post("https://pragnakalp-ocr-image-to-text.hf.space/run/predict", json={
        "data": [
            "PaddleOCR",
            f"data:image/png;base64,{image_base64}",
        ]
    }).json()

    data = response.get("data", [])

    text = " ".join(data)  
    audio_file = trans(text, lang)

    return audio_file


def operator(image_array, value, lang):
    if value == "1":
        audio_file = object_recognition(image_array, lang)
    elif value == "2":
        audio_file = ocr_detection(image_array, lang)
    else:
        text = "Sorry, I can't perform this operation."
        audio_file = trans(text, lang)
    return audio_file

# Create Gradio interface
iface = gr.Interface(fn=operator, inputs=["image", "text", "text"], outputs="audio")
iface.launch(share=True)